Calculation of Deadline Missing Probability in a QoS Capable Cluster Interconnect
Infinite chain of N different deltas a simple model for a Quantum Wire
Infinite chain of N different deltas: A simple model for a quantum wire
Jose M. Cerver´ o∗ and Alberto Rodr´ ıguez
PACS Numbers: 03.65.-w: Quantum Mechanics 71.23.An: Theories and Models; Localized States 73.21.Hb: Quantum Wires
∗
cervero@al.es. Author to whom all correspondence should be addressed
3 model. We close with a Section of Conclusions.
1
Periodic Array
Let us consider an electron in a periodic one dimensional chain of atoms modelled by the potential constituted by an array of N delta functions each one with its own coupling e2 i, (i = 1, 2, ...N ). After finishing the N -array, the structure repeats itself an infinite number of times. The number of species N , can be arbitrarily large but finite. The case N =1 is an old textbook exercise but may be convenient to be revisited [19] for taking a full profit of our general results. The generalization can thus be followed in a more straightforward manner. The relevant primitive cell for N = 2 can be represented for the following set of wavefunctions:
Performance Analysis of Probabilistic Flooding Using Random Graphs
Performance Analysis of Probabilistic Flooding Using Random GraphsKonstantinos Oikonomou Dept.of Informatics,Ionian University Tsirigoti Square7,49100Corfu,Greeceokon@ionio.grIoannis StavrakakisDept.of Inform.&Telecoms,Univ.of Athens Panepistimiopolis,Ilissia,15784,Athens,Greeceioannis@di.uoa.grAbstractProbabilisticflooding(parameterized by a forwarding probability)has frequently been considered in the past,as a means of limiting the large message overhead associated with traditional(full)flooding approaches that are used to disseminate globally information in unstructured peer-to-peer and other networks.A key challenge in using proba-bilisticflooding is the determination of the forwarding prob-ability so that global network outreach is achieved while keeping the message overhead as low as possible.By showing that a probabilisticflooding network gener-ated by applying probabilisticflooding to a connected ran-dom graph network can be bounded by properly parame-terized random graph networks and invoking random graph theory results,bounds on the value of the forwarding prob-ability are derived guaranteeing global network outreach with high probability,while significantly reducing the mes-sage overhead.Bounds on the average number of messages –as well as asymptotic expressions-and on the average time required to complete network outreach are also de-rived,illustrating the benefits of the properly parameterized probabilisticflooding scheme.1.IntroductionIn modern network architectures such as peer-to-peer networks,global node outreach(i.e.,reaching all network nodes)is a major challenge.Reaching all nodes in a net-work is frequently required either to disseminate informa-tion(e.g.,advertise a certain service)or retrieve informa-tion(e.g.,service discovery).In structured peer-to-peer net-works,the available structure facilitates the global network outreach or reaching the appropriate node with relatively low delay and message overhead,[17],[4],[22],[15],[2], [20].In unstructured peer-to-peer networks,though,(e.g. Gnutella,[14]),the global network outreach is far more challenging to achieve efficiently,as there is no structure to take advantage of and design an effective scheme.As a result,the brute-force approach is followed,typically imple-mented through resource wasteful approaches such asflood-ing,[14],[9],[8],[10].Traditionalflooding that traverses all network links and reaches all network nodes,is not an efficient approach as it requires a number of messages equal to the number of network links.In view of the typically large size of peer-to-peer networks in terms of nodes and links,it is clear that traditionalflooding would not be effective for such envi-ronments.However,flooding is frequently considered for comparison purposes and in order to establish the relative efficiency of alternative schemes.Many variations of traditionalflooding have been pro-posed for service discovery in unstructured environments. In Gnutella,[14],a TTL(Time-To-Live)value is used to restrict messageflooding to a small number of hops around the node that has initiated the searching process(this node will be referred to as the initiator node).This approach may be scalable for small values of TTL but at the same time it significantly reduces the probability of locating the requested node(s)of interest in large peer-to-peer networks.Random walks,e.g.[9],[18],have been proposed to reduce the total number of messages by sending a limited number of special messages(agents)in the network.Each of them follows its own path by choosing randomly the next hop node.Messages terminate their walk either af-ter some time(e.g.,TTL expiration)or after checking with the initiator node and learning that the node of interest has already been discovered by another message,or a combi-nation of both.Hybrid probabilistic schemes(e.g.,a local flooding process initiated after a random walk)have also been proposed and analyzed,[8],as well as other schemes that adapt the employed TTL values in a probabilistic man-ner,[10].Another modification,[24],allows for network nodes to forward messages to their neighbors in a random manner,thus significantly reducing the number of messages in the network.The aforementioned idea of reducing the messages of traditionalflooding by selectively choosing the next hop nodes,lays also behind probabilisticflooding,[6],[26],[12],[23].Under probabilisticflooding,messages are forwarded to neighbor nodes based on a certain forwarding probability.There is clearly a trade-off between the induced total number of messages and the number of nodes that are actually reached by such messages:the smaller the prob-ability the smaller the message overhead and the larger the set of nodes in the network not being accessed through these messages.The work presented here investigates probabilisticflood-ing when the underlying network is a random graph,[13], and aims at designing such a scheme in a way that the afore-mentioned trade-off is well managed.That is,achieve high node reachability with a relatively small number of mes-sages.Analytical tools and results,borrowed from ran-dom graph theory,[13],[7],are considered for analyzing the probabilisticflooding.One of the main contributions of this work is establishing a connection between random graphs and the probabilisticflooding network;the latter is defined to be the network consisting of the(sub)set of links and nodes of the underlying random graph network that are traversed by the messages under the probabilisticflooding.It should be noted that the idea of randomly choosing the next neighbor node is not new and has been the subject of many research works in the past,[6],[26],[12],[23], [5],[25],[16],[21],[1],[27].Even though many of these works are related to probabilisticflooding(e.g.,[6],[26], [12],[23]),none of these works have addressed the prob-lem of deriving analytically boundaries f or the value of the forwarding probability that achieves global node outreach for a random graph.In fact,another contribution of this work is the derivation of analytical bounds on the appro-priate value of the forwarding probability,defined to be the value for which the probabilisticflooding network contains (with high probability)all network nodes(i.e.,all nodes are reached)using the smallest possible number of messages. To the best of the authors’knowledge this is thefirst time that such a result is derived for the particular environment. Equally important is the use of random graph theory for the analysis of a particular algorithm(i.e.,probabilisticflood-ing),as it may trigger more such considerations in this re-search area and facilitate the study of information dissemi-nation schemes under a new perspective.Finally,another contribution of this work is the deriva-tion of an upper bound on the(average)total number of messages under the probabilisticflooding.It turns out that this number(under the appropriate value of the forwarding probability)is significantly smaller than that induced under the traditionalflooding.However,as it is analytically shown in this paper,the price paid for this reduction is(a)an in-crease of the time required to outreach all network nodes (global outreach time);(b)global node outreach is achieved with high probability as opposed to certainty under tradi-tionalflooding.Section2summarizes important results from random graph theory that will be used throughout this work.Section3presents the probabilisticflooding scheme and Section4 discusses its connection to the random graphs.Analytical results are presented in Section5and conclusions are drawn in Section6.2.The Random Graph ModelUsually a network is represented by a graph G(V,L), where V is the set of nodes and L is the set of(bidirectional) links connecting the nodes.For example,if a link(u,v), exists between node u and node v,then(u,v)∈L.Ran-dom graphs,mainly introduced by the pioneering work of P.Erd˝o s and A.R´e nyi,[13],have some properties that help to shed light on various aspects of networks.These prop-erties appear in many different networks:social contacts, biological networks,telecommunication networks etc.,[3], [19].In the sequel,a random graph(and the correspond-ing network)will be represented by G p(N),where N is the number of nodes in the network and p an independent prob-ability that a link exists among any pair of network nodes, [7].For most of the cases,as it is also the case in this work, N is considered to be significantly large.A simple construction model to create a G p(N)network,[7],[3],[19],is to consider at the beginning only one node present in the network(e.g.,node0)and assume that nodes entering the network at any order(e.g.,node1entersfirst followed by node2etc.)follow the next rule:each node ar-riving in the network creates a link with any of the already existing nodes with probability p and it does not create the particular link with probability1−p.Consequently,when node1enters the network(only node0is present)a link is created(or not)with probability p(or1−p).When node2 enters the network(and nodes0and1are already present) a link is created(or not)between node2and node0with probability p(or1−p)and another link is created(or not) between node2and node1with probability p(or1−p). By the time the N-th node enters the network,there will beon average pN N−12links in G p(N),[7].From the afore-mentioned construction process,it is evident that for p=0, there are no links in the resulting graph,whereas for p=1, the resulting graph is the complete graph(i.e.,it contains all possible links among the N nodes that amount to N N−12).At this point it is important to note that in most of the cases the arguments are made with high probability(w.h.p.).For example,for p<1N, P r[the giant component exists]→0,while for p>1N, P r[the giant component exists]→1,[18].Actually,for p=1N,where the shape of the network suddenly changes, a phase transition,[7],phenomenon takes place.For p=log(N)Nall nodes become part of the giant component and the network becomes connected w.h.p.,[7].Thus,for any value of p ≥log(N )N ,G p (N )is connected w.h.p.The average number of links,|L |,for the network corre-sponding to G p (N ),when p =log(N )N ,[7],is given by,|L |=12(N −1)log(N ).The diameter of the resulting connected network,denoted by D ,has been proved,to“concentrate around”,[3],log(N )log(pN ),[11],which allows for,D ≈log(N )log(pN ).3.Probabilistic FloodingUnder probabilistic flooding,[6],the initiator node sends a message to each of its neighbor nodes with an (indepen-dent)forwarding probability p f .Any node receiving such a message forwards it to each of its own neighbor nodes (except from the node the message arrived from)with prob-ability p f .Clearly,for p f =0,no messages are sent in the network,while for p f =1,probabilistic flooding re-duces to traditional flooding.As a result of the probabilistic flooding,a network can be defined that consists of the set of nodes that have been reached by the messages and the set of links over which these messages have been forwarded.This particular network will be referred to hereafter as the prob-abilistic flooding network .It is easy to show (based on the definition of the probabilistic flooding)that the probabilistic flooding network is actually a connected network each link of which corresponds to exactly one forwarded message.The main objective in this paper is to derive appropriate values of the forwarding probability that will yield a proba-bilistic flooding network that will include all network nodes (i.e.,all nodes will be reached under probabilistic flooding)w.h.p.and at the same time -the average number of links contained in this probabilistic flooding network be as small as possible (to keep the (average)total number of messages small).Consider a connected (i.e.,p ≥log(N )N ))random graph network G p (N )as the underlying network.Let F p f (G p (N ))denote the probabilistic flooding network gen-erated over the (random graph)network G p (N )when prob-abilistic flooding is employed with probability p f .Under probabilistic flooding a message is forwarded with proba-bility p f over each of the links of G p (N )that are attached to it (except from the link from which the message arrived).As a link connects two different nodes and these nodes may receive a message through a different link (one of the other links attached to them),it is possible that both nodes at-tempt a message transmission over this common link (at the same or at different times).This will happen,for example,if one of the nodes receives a message first through a dif-ferent link,this node makes a failed attempt to forward a message over the common link (with probability (1−p f )),the other node receives a message (through another link)and consequently attempts a message forwarding over thecommon link (again,with probability (1−p f )).In such cases,a link will have two opportunities to forward a mes-sage and,thus,become part of F p f (G p (N )).Other links will have only one opportunity,though;for instance,this will be the case when a message forwarding attempt over a link fails and the node at the other end of the link never receives a message through one of its other links.Links of G p (N )which have only one opportunity to forward a message will be included in F p f (G p (N ))with probability p f ,while links of G p (N )which have two opportunities to forward a message will be included in F p f (G p (N ))with probability 1−(1−p f )(1−p f )=2p f −p 2f (to simplifythe notation let ˜p =2p f −p 2f ).By construction,the resulting probabilistic flooding net-work over a connected G p (N )network (F p f (G p (N )))seems to have a certain resemblance to random graphs.Such observations -allowing for the use of random graph theory for the analysis of probabilistic flooding -are dis-cussed and taken advantage of in the following section.4.Random Graph Network Representation of Probabilistic FloodingGiven that for each node of a connected network is as-sociated with at least one link and most likely with several,removing a link from a network does not necessarily dis-connect (or remove)an associated node as well.In other words,the decrease in the number of nodes in a network as a result of a decrease in the number of links is expected to be lower than the decrease in the number of links.Conse-quently,it is conceivable that all network nodes continue to be included in a network (i.e.,be connected)with high prob-ability (w.h.p.)despite the removal of a number of links.This observation suggests that a probabilistic flooding net-work with sufficiently high forwarding probability may still keep all the nodes connected and in the network,despite a potentially significant removal of links due to a decision not to forward a message over such links.In view of the above discussion it is evident that as p f decreases,the number of links in F p f (G p (N ))decreases as well,while the number of nodes in F p f (G p (N ))decreases at a lower rate.Consequently,for a small reduction in p f below the value of 1,it is expected that all network nodes be still included in F p f (G p (N ))w.h.p.It is thus expected that there is a certain value for the forwarding probability,denoted by p f,0,such that:(a)if p f <p f,0,then the proba-bilistic flooding network does not include all network nodes w.h.p.;(b)if p f ≥p f,0,then the probabilistic flooding net-work does include all network nodes w.h.p.p f,0will be referred to as the appropriate value of the forwarding prob-ability.The determination of p f,0is not an easy task and the focus in the sequel is on the analytical derivation of up-per and lower bounds.First consider the G p ×p f (N )ran-dom graph.G p ×p f (N )can be constructed using the con-struction model presented in Section 2or simply consid-ering G p (N )and then independently selecting each link of G p (N )with probability p f .Keeping in mind that the probabilistic flooding network is created by independently selecting links from G p (N )with probability p f for some of them and with probability ˜p for some others,it is evi-dent that F p f (G p (N ))contains on average more links than G p ×p f (N ).Consequently,when G p ×p f (N )is connected w.h.p.,then F p f (G p (N ))is also connected w.h.p.and,thus,includes all network nodes w.h.p.Before proceeding it is in-teresting to see whether F p f (G p (N ))can have (on average)as many links as G p ×p f (N ).This will be true when all the links of G p (N )are selected with the same probability p f under probabilistic flooding.This is the case,for example,when G p (N )is actually a tree and consequently,all links are selected with the same probability p f .A second observation is possible between F p f (G p (N ))and G p טp (N ).G p טp (N )can be created by independentlyselecting links from G p (N )with the same probability ˜p.Clearly,G p טp (N )contains (on average)more links than G p ×p f (N )(note that p f ≤2p f −p 2f ;the equality hold-ing for p f =1)and when the latter network is con-nected the former is also connected w.h.p.However,in contrast to the previous case,F p f (G p (N ))may not (on average)be as dense as G p טp (N ).This is easily con-cluded since under probabilistic flooding there exists at least one link that has been selected with probability p f .Actually,there are more than one:all links over which messages have been forwarded for the first time to a par-ticular node (e.g.,from the initiator node to its neighbor nodes).Consequently,F p f (G p (N ))contains (on average)fewer links than G p טp (N ).From the previous observa-tion it is now possible to derive analytical bounds for p f,0.From the discussion presented in Section 2,random graphG p ×p f (N )becomes connected w.h.p.for pp f =log(N )N ,or p f =log(N )pN .G p טp(N )becomes connected w.h.p.for p ˜p =log(N )N ,or p (2p f −p 2f )=log(N )N ,or pNp 2f −2pNp f +log(N )=0.The latter polynomial has two solutionsfor p f ,p f,1−2=2pN ±√4p 2N 2−4pN log(N )2pN ,or p f,1−2=1±1−log(N )pN .Note that since p ≥log(N )N(G p (N )is connected w.h.p.),1−log(N )pN ≥0.The solution that satis-fies 0≤p f ≤1,is given by p f,1=1−1−log(N )pN .As p f increases and by the time G p ×p f (N )becomesconnected w.h.p.,F p f (G p (N ))has already become con-nected as well and includes (on average)all network nodes w.h.p.since it contains on average more links than G p ×p f (N )(see earlier).Therefore,it is evident that,p f,0≤log(N )pNis satisfied (the equality holds only for the particular case that all links under probabilistic floodingare selected with probability p f ).On the other hand,ran-dom graph G p טp (N )contains more links (on average)than F p f (G p (N ))and therefore,it becomes connected w.h.p.forsmaller values of p f .Therefore,p f,0>1−1−log(N )pN .Eventually,1−1−log(N )pN <p f,0≤log(N )pN .(1)The latter inequality is an important one since it boundsthe particular value of the forwarding probability p f for which all network nodes are reachable w.h.p.with the smallest possible number of messages.Actually,the ob-servation that F p f (G p (N ))“lays”between G p ×p f (N )and G p טp (N )allows for (a)the use of the aforementioned “safe”value for p f to ensure that all network nodes are reached;(b)to derive analytical upper bounds (i.e.,worst case scenarios)with respect to the total number of messages and termination time.The corresponding analysis is the fo-cus of the following section.5.AnalysisFor comparison reasons,traditional flooding is also con-sidered here.Given that the number of edges of a randomgraph is on average pN N −12,the average number of mes-sages under traditional flooding,M t ,is given by,M t =pN N −12.Since for p f =log(N )pN ,F p f (G p (N ))contains allnetwork nodes G p (N ))w.h.p.,it is useful to derive the cor-responding average number of messages,M p .Clearly,M p corresponds to the average number of network links and it is already known that M p lays (on average)between the average number of links of G p ×p f (N )and G p טp (N ).The average number of links for G p ×p f (N )(G p טp (N ))is equalto pp f N N −12(p ˜p N N −12)and for p f =log(N )pN this number becomes 12(N −1)log(N )((1−log(N )2pN )(N−1)log(N )).Eventually,12(N −1)log(N )≤M p <(1−log(N )2pN)(N −1)log(N ).(2)From Equation (2)it is clear that M p increases with N as N log(N ),while M t increases as N 2.Consequently,the probabilistic flooding can reduce the message overhead substantially when N is large.Let R M =M upM t,where M updenotes the upper bound of the average number of messagesunder probabilistic flooding (i.e.,M u p =(1−log(N )2pN )(N −1)log(N )).Eventually,R M =2log(N )pN − log(N )pN2.(3)As expected from the previous discussion,R M →0,as N →+∞.Figure 1.a illustrates the message overhead savings by plotting R M versus N (100≤N ≤2000)and for various values of p .Note that throughout this work N was considered to be significantly large.and this is also the case under which most of the For in-stance,for a network G 0.8(1000),global network outreach can be achieved under probabilistic flooding with p f in the range (0.00188,0.00375)–see Equation (1)–with only around 1%of the messages expected under traditional flooding.The overhead savings of the probabilistic flood-ing are gained with the following costs.First,the global network outreach achieved under traditional flooding with certainty is now achieved only with high probability.That is,global outreach is only probabilistically guaranteed.a. b.Figure 1.R M and R t as a function of N for various values of p .The second cost paid for the messages overhead savings is regarding the time to complete the global outreach.The average maximum such time (assuming that the initiator is located at the network boundaries)is equal to the networkdiameter.Thus,t t =log(N )log(pN ),under the traditional floodingsince the diameter of G p (N )is equal to log(N )log(pN ).Since the diameter of F p f (G p (N ))lays between those of G p ×p f (N )and G p טp (N )w.h.p.,the average global outreach time t p ,is bounded as follows,log(N )log(p ˜p N )<t p ≤log(N )log(pp f N ).(4)Let t u p denote the upper bound in Equation (4)and let R t =t u pt t.Eventually,R t =log(pN )log(log(N )).(5)Since the diameter of F p f (G p (N ))can never be smaller than that of G p (N ),R t ≥1;equality holds when the two diameters are equal in which case the global network out-reach time are also equal.R t shows the time required to achieve global outreach in F p f (G p (N ))as a percentage ofthat in G p (N ).Figure 1.b illustrates R t as a function of N and for various values of p .For instance,for G 0.8(1000),the global network outreach time under probabilistic flood-ing is about 3.5times that under traditional flooding.Re-call for Figure 1.a that for the same example,only 1%of the total number of messages under traditional flooding were needed (on average)for global network outreach un-der probabilistic flooding.6.ConclusionsIn this work,the problem of limited information dissem-ination in large,unstructured networks is considered and specifically the focus has been on schemes that can achieve a global network outreach.Global network outreach is available in a network if the employed information dissem-ination scheme is capable of taking a message from any originating node to any other network node.Such network outreach is needed in order to support routing protocols,ad-vertise a new service,search for some information,etc.Tra-ditional flooding schemes achieve global network outreach in unstructured networks with certainty (deterministically,for a connected network)at a large message overhead cost.In this paper,probabilistic flooding schemes have been con-sidered in order to reduce their associated large overhead,at the price of providing probabilistic global network outreach guarantees.It is shown here that the network created by probabilis-tic flooding over a random graph network lays between two random graph networks -which are determined facilitating this way the derivation of analytical bounds on the value of the forwarding probability that results in a fairly decreased (compared to the traditional flooding)number of messages,while achieving global node outreach with high probabil-ity (w.h.p.).In particular,it is shown that the probabilistic flooding network F p f (G p (N ))generated by applying some forwarding probability p f over a connected random graph underlying network G p (N ),“lays”between two random graph networks:G p ×p f (N )and G p טp (N ).Actually,the probabilistic flooding network F p f (G p (N ))includes (on average)at least as many links as in G p ×p f (N )and at most as many as in G p טp (N ).If G p ×p f (N )is a connected net-work w.h.p.,then the probabilistic flooding network con-tains all network nodes w.h.p.The latter observation leads to the determination of a “safe”value for the forwarding probability p f that ensures w.h.p.that all nodes are reached under probabilistic flooding,even though some extra mes-sages might be required compared to the case of the optimal value that is hard to determine though.A comparison of the probabilistic flooding under the safe forwarding probability with the traditional flooding is car-ried out.It is shown that the number of messages under probabilistic flooding increases as N log(N )as opposed toN2under traditionalflooding.Thus,significant message overhead reduction can be achieved,especially for large networks(large N).The relative message overhead(com-pared to that under traditionalflooding)is also derived and shown to yield substantial message overhead savings even for low to medium values of N.However,as it was men-tioned,the global network outreach achieved under tradi-tionalflooding with certainty is now achieved only with high probability.Finally,the increase in the time needed to get the message across the network is also derived un-der both schemes and compared.As expected,this time is slightly higher under probabilisticflooding(compared to message reduction).Although theoretical results from random graphs support the claim in various places of this paper for results w.h.p.,it would be useful to assess this theoretical statements by sim-ulating connected random graph networks of various sizes and measuring the node outreach percentile;that is,mea-sure the probability that at least x%of the nodes are in-cluded in the probabilisticflooding network and are,thus, reachable.7.AcknowledgmentsThis work has been supported in part by the project Au-tonomic Network Architecture(ANA)(IST-27489),which is funded by IST FET Program of the European Commis-sion.References[1] D.T.A.Ganesh,L.Massoulie.The effect of network topol-ogy on the spread of epidemicss.In13-17March2005,IN-FOCOM2005,volume2,pages1455–1466,2005.[2]J.K.AB.Y.Zhao and A.Joseph.Tapestry:An infrastructurefor fault-tolerant wide-area location and routing.In Tech.Rep.UCB/CSD-01-1141,UCB,2001.[3]R.Albert and A.Barab´a si.[4] B.Awerbuch and D.Peleg.Concurrent online tracking ofmobile users.In ACM SIGCOMM Symposium on Commu-nication Architectures and Protocols,1991.[5]T.C.B.Williams,D.Mehta and W.Navidi.Predictive mod-els to rebroadcast in mobile ad hoc networks.IEEE Trans-actions on Mobile Computing(TMC),3:295–303,2004. [6] F.Banaei-Kashani and C.Shahabi.Criticality-based analy-sis and design of unstructured peer-to-peer network as“com-plex systems.In in Proceedings of the Third International Symposium on Cluster Computing and the Grid,pages351–358,2003.[7] B.Bollob´a s.Random Graphs.Cambridge University Press,second edition edition,1999.[8]M.M.C.Gkantsidis and A.Saberi.Hybrid search schemesfor unstructured peer-to-peer networks.In IEEE Infocom 2005,2005.[9] E.C.K.L.C.Lv,P.Cao and S.Shenker.Search and repli-cation in unstructured peer-to-peer networks.In ICS2002, 2002.[10]N.B.Chang and M.Liu.Optimal controlledflooding searchin a large wireless network.In Third International Sympo-sium on Modeling and Optimization in Mobile,Ad Hoc,and Wireless Networks(WiOpt’05),pages229–237,2005. [11] F.Chung and L.Lu.The diameter of sparse random graphs.Adv.in Appl.Math.,26(4):257–279,2001.[12]V.V.Dimakopoulos and E.Pitoura.On the performanceofflooding-based resource discovery.EEE Transactions on Parallel and Distributed Systems,17(11):290–297,Novem-ber2006.[13]P.Erd˝o s and A.R´e nyi.On random graphs.PublicationesMathematicae Debrecen6,pages290–297.[14]Gnutella.Gnutella rfc.http://rfc-gnutella./,2002.[15] D.K.M.K.I.Stoica,R.Morris and H.Balakrishnan.Chord:A scalable peer-to-peer lookup service for internet applica-tions.In in Proceedings of ACM SIGCOMM’01,2001. [16]R.Karp,C.Schindelhauer,S.Shenker,and B.V ocking.Randomized rumor spreading.In Proceedings of the41st Annual Symposium on Foundations of Computer Science, FOCS’00,2000.[17] F.H.M.van Steen and A.Tanenbaum.A model for world-wide tracking of distributed objects.In TINA’96Conference, 1999.[18]P.V.Mienghem.Performance Analysis of CommunicationNetworks and Systems.Cambridge University Press,2006.[19]M.E.J.Newman.The structure and function of complexnetworks.SIAM Review,45:167–256,2003.[20] A.Rowstron and P.Druschel.Pastry:Scalable,distributedobject location and routing for large-scale peer-to-peer sys-tems.In in IFIP/ACM International Conference on Dis-tributed Systems Platforms(Middleware),pages329–350, 2001.[21] A.G.B.P.S,Boyd and S.D.Randomized gossip algo-rithms.IEEE/ACM w.,14:2508–2530,2006. [22]M.H.R.K.S.Ratnasamy,P.Francis and S.Shenker.Ascalable content-addressable network.In in Proceedings of ACM SIGCOMM’01,2001.[23] D.Tsoumakos and N.Roussopoulos.Adaptive probabilisticsearch for peer-to-peer networks.In3rd IEEE International Conference on P2P Computing,2003.[24] D.G.V.Kalogeraki and D.Zeinalipour-Yazti.A local searchmechanism for peer-to-peer networks.In in CIKM(Interna-tional Conference on Information and Knowledge Manage-ment),2002.[25] B.Williams and parison of broadcasting tech-niques for mobile ad hoc networks.In ACMSymposium on Mobile Ad Hoc Networking and Computing(MOBIHOC), pages194–205,2002.[26] D.C.Yoav Sasson and A.Schiper.Probabilistic broadcastforflooding in wireless mobile ad hoc networks.In Swiss Federal Institute of Technology(EPFL),Technical Report IC/2002/54.[27]J.Y.H.Z.J.Haas and L.Li.Gossip-based ad hoc routing.IEEE/ACM w.,14:479–491,2006.。
最新版本工程应用英语
工程应用英语课程一．单选题：1. Computers are(B. useless)unless they are given clear and accurate instructions and information.2. Active (B. Recruiting) for engineers often begins before the student’s last year in the university.3. For the student who is preparing to become a (C. Civilengineer), these specialized courses may deal with such subjects as geodetic surveying, soil mechanics.4. The civil engineer may work in research, design, construction,(B. Supervision), maintenance, or even in sales.5. Civil engineers work on many different kinds of (C. Structures).6. In designing buildings, civil engineers often work as (B. Consultants) to architectural or construction firms.7. Dams, bridges and other large projects ordinarily employ several engineers whose work is coordinated by a (D. Systems) engineer who is in charge of the entire project.8. Construction is a(B. Complicated) process on almost all engineering projects.9. In compression, the material is (C. Pushed) together.10. When a saw cuts easily through a piece of wood, the wood is (A. in tension).11. We defined (D. Shear) as the tendency of a material to fracture along the lines of stress.12. The principal construction materials of earlier times were wood and (C. masonry brick),stone, or tile, and similar materials.13. Modern cement is a mixture of(B. limestone and clay).14. Concrete is very (D. Versatile) 15. Steel has great tensile strength whileconcrete has great compressive strength,thus, the two substances (C. Complement)each other.16. One system that helps (A. Cut)concrete weight to some extent usespolymers.17. The retention money serves to insure(D. the employer)against any defects thatmay arise in the work.18. The civil engineering work must becompleted to the satisfaction of theemployer, or his(D. Representative)19. For moderate and longer hauls,self-loading scrapers pulled byrubber-tired hauling units and push-loadedby tractors offer(B. Lower) cost.20. Highway maintenance activities canbe grouped and classified according (D.to)the purpose of the treatment.21. Engineering is a profession, whichmeans that an engineer must have aspecialized(D．university )education.22. In most cases, the tender maybe(B．Withdrawn)at any time until it hasbeen accepted.23. Current trend is to require students totake courses in the(C.social)science andthe language arts.24. The law relating to contracts imposeson each party to a contract (D. a legalobligation) to perform.25. Indeed, the civil engineer’s choiceis(C. large)and varied.26. Roadbeds (B. underlie) highwaypavement structures and the ballast andtrack on which trains move.27. Construction can be very (A.dangerous).28. Where material is moved less thanabout 60m or steeply downhill, driftingwith a track or wheel type bulldozer is (A.cheapest)29. Thrust is the pressure exerted by eachpart of a structure on (A. its other part)30. The weight of all the people, cars,furniture, and machines and so on that thestructure will support when it is in use is(B. live load)31. In tension, the material is(B. pulledapart)32. In fills constructed by end dumping orby placing in thick layers, material,density, and moisture content could (C.vary) greatly from one spot to another.33. Both (A. steel and cement), the twomost important construction materials ofmodern times, were introduced in thenineteenth century.34. The total station is used to measureangles in both vertical and horizontalplanes, and the level to measure (C.elevation difference)35. The (B. defective )vehicle is a creatorof accidents.36. Prestressed concrete is an (B.improved )form of reinforcement.37.A simple contract consists of anagreement entered into by( D. two ormore parties)38.(B. V olume) change would result indifferential settlement or swell betweenadjacent areas.39. There are two basic procedures forcontrolling the embankment density:‘manner and method’ and ‘(A. result )’.40.A main source of accidents, theproblem of(C. drunken)driving is the mostserious of all.41. Computer programming is nowincluded in almost all engineering(B.curricula).42. The relationship between engineeringand society is getting( C. Closer)43. Types of contracts are virtually classified by their(D. payment)system: (1) price-based and (2) cost-based.44. Computers can’t solve complicated problems unless they are given( D. a good program)45. In recent years, rippers have been used successfully to( C. break up)loose or fractured rock.46. Civil engineering projects are almost always (A. distinctive)47. Usually there are (C. no)easy answers on equipment selection.48. (A. Vertical)force acts up and down.49. Layered construction also produced greater uniformity in the material( D. itself)and in its density and moisture content.50. The actual cost of any single highway traffic accident is extremely( C. difficult )to determine.51. Basically, causes of automobile accidents can be categorized(D. into)four major groups.52. Electronic distance measuring (EDM) not only can measure the distance between objects but also determine( A. the direction)53.(A. Two of )the recent improvement in visibility are wraparound windshields and narrowed roof support pillars.54. There is a great deal that the actual highway designer can( C. do to)prevent accident.55. To avoid the driving after drinking, one of the methods is( B. breath test ). 56. It is suitable for remote sensing technique to be used for highway location in(C. mountainous country without forest).57. The information on the aerial photographs can be converted into mapswith the help of stereoscopes which isable to see objects in(C. three dimensions)58. The normal steel does not exert anyforce of its own on the member,( B.contrary)to the action of prestressingsteel.59. The extensive use of prestressedstructures has been due in ( A. no )smallmeasures to the advances in thetechnology.60.The employer selects the contractor forthe project by( D. Bidding).61. Many different( A. corporations ) andgovernment agencies have competed forthe services of engineers in recent years.62. Civil engineers may prefer to workwith one of the government agenciesthat( B. deals )with water resources.63. It is normal practice for( B. a)engineerto specialize in just one kind.64. Construction involves the work andutilizing the equipment and the materialsso that costs are kept as( C. low )aspossible.65. For example, (B. dams)are often builtin wild river valleys or gorges.66. Electrical and mechanical engineerswork on the(A. design )of the powerhouseand its equipment.67. In Rome, most of the people livedin(B. insulse ), great tenement blocks thatwere often ten stories high.68. The prospective civil engineer shouldbe aware of the physical( A.Conditions)that will be made on him orher.69. Much of the work of civil engineeringis carried on( C. outdoors)70. In addition, the building ofskyscrapers, bridges and tunnels must alsoprogress under all kinds of(C.weather)conditions.71. The Romans also used a naturalcement called pozzolana, made from (B.volcanic ash), that became as hard asstone under water.72. Different proportions of theingredients produce concrete with( A.different )strength and weight.73. (A. Prestressed) concrete has made itpossible to develop buildings with unusualshapes.74. The modern engineer must alsounderstand the (C. different)stresses towhich the materials in a structure aresubject.75. Today, scientific data permit theengineer to make careful calculations( D.in advance)76. The force which the live load will beexerted on the structure is( C. Impact)77. When a saw begins to bind, the woodis( A. in compression because)the fibersin it are being pushed together.78. ( D. Steel )rods are bent into theshapes to give them the necessary degreeof tensile strength.79. Many great buildings built in earlierages are massive structures with( B. thickstone walls)80. We all enter into contracts almostevery day for the supply f goods,(Btransportation)etc.81. Some contracts must be made in aparticular(D. form)to be enforceable.82. Once a person has signed a documenthe is assumed to have(B. approved)itscontents.83. By setting down the terms of acontract in writing one secures avoiding( A. disputes)84. In an entire contract, where( D. theemployer)agrees to pay a certain sum inreturn for civil engineering work..85. (B. The contractor ) is not entitled to any payment if he abandons the work prior to completion.86. The contractor is not entitled to receive payment in (A. full )until the work is satisfactorily completed.87. A tender is normally required to be a definite( C. offer)88. Generally, civil engineering contracts provide for the issue of (B. interim certificates)at various stages of the works.89. It does not give the employer the right to demand an(A. unusually)high standard of quality throughout the works.90. The employer does not usually bind himself to accept the lowest or indeed any tender and this is often stated in the(C. advertisement )91. A contract has been defined as an agreement which directly creates and contemplates( C. an obligation)92. When we enter into contracts we are willing to(C. pay )for the service we receive.93. If there is no written agreement and( C.a dispute )arises in respect of the contract.94. The rubber-tired tractor units have difficulty in operating on( D. wet), slippery roadbeds.95. There are( A. Many)variables in earthmoving.96. The term(D. Embankment)describes the fill added above the low points along the roadway to raise the level to the bottom of the pavement structure.97. Material for( B. embankment)commonly comes from roadway cuts or designated borrow areas.98.(C. Field) control is largely a matter of conducting the specified procedure.99. Modern practice requires that embankment construction be( A.carefully )executed and controlled.100. Construction of pavement over highfills often was( B. deferred )for a year ormore after completion of the fill to allowthe settlement to occur.101. Nearly( B. all)vegetable mattershould be removed from the originalground and fill material.102. A track or wheel type bulldozer is( D.not suitable)to earthmoving ofconsiderably long hauls.103. Loose rock is handled by( A.tractor-scraper)units as is done with‘common’ excavation.104. The highway can require mentaland( A. physical) response.105. The needs generated by the greatincrease( D. in)vehicle numbers andkilometers of road have given rise tomajor research programs in trafficplanning.106. Terminology concerned( B.with)highway preservation variesconsiderably from country to country.107. Highway improvement is also a keyfactor( B. in) preventing accidents.108. The actual degree of safety oneexperiences on a given highway isdetermined by decisions made on ( B.different) levels.109. Public agencies typically dictate themajor constraints within which thesedesign decisions are( A. to be) made.110. Finally, individual motorists makedecisions regarding their own safety ( C.as) they select speed, route for their cars.111. Safe highways are ( C. expensive)and it appears that the driving public doesnot want safe highways.112. People do not want to pay the costsand suffer the restrictions necessary toproduce ( A. safety) in traffic.113. It is often ( A. impossible)todetermine the true condition of a vehicleafter a crash.114. No figures( B. are)available tojustify it.115. For the driver’s vision, in the body ofthe automobile, both side and rearwindows have been greatly( D. enlarged)in area.116. Another improvement in drivervisibility is the introduction of theremote-controlled( B. outside)rearviewmirror.117. The safe performance of the brakesystem( C. under)high temperatures hasbeen ensured.118. Relocation and reduction in theheight of the brake pedal has meant thatthe brake can be applied( A. much)rapidly.119. The use of uniform traffic controldevices will reduce driver reactiontime( A. as well as) confusion.120. Removal, relocation and redesign offixed obstructions, can provide a clear( C.recovery) area for vehicles out of control.121. Vehicular safety design usuallycenters( B. upon) protecting the driver andhis passengers.122. The highway construction may alsocause( D. adverse)impacts on thesurroundings.123. The designed highway alignmentmust meet the technical( B. standard)ofthe highway engineering.124.( A. Hot rolled asphalt)is a gapgraded material with less coarseaggregate.125. In this case, layer thickness, moisturecontrol, and the number of passes by aroller of specified type and weight are( A.predetermined).126.( B. Ground survey)is the conventional location technique for highway.127. A( B. total station)is only used for measuring the vertical heights of objects. 128. If Party A commissions Party B to execute the construction work, then Party B is referred to as( B. the contractor). 129. The force-account work should be checked and approved daily by( D. both A and B) .130.( A. Fast speed)is not advantage of highway transportation.二．填空题：131. Engineers often work as( consultants)to architectural or construction firms.132. Young engineers may choose to go into( environmental)or sanitary engineering.133. It is sufficient in order to create a legally( binding), if the parties express their agreement and intention to enter into such a contract.134. One party to the contract is( liable)for breach of contract if he fails to perform his part of the agreement. 135.( Clearing)the site precedes all grading and most other construction operations.136. Loose rock includes materials such as( rotten or weathered)rock, or earth mixed with boulders.137. No attempt was made to control( moisture)content or to secure compaction.138. The( redesign)of windshield wipers, fresh air ventilating systems, had result in greater vehicle safety.139. The safe performance of the brake system has been ensured by the use of( heavy-duty)brake fluid.140. Relocation and reduction in height ofthe brake( pedal)has meant that thedriver’s total reaction time has beenreduced.141. Areas of research connected withcivil engineering include soil mechanicsand ( soil stabilization) techniques.142. Modern cement, called ( Portlandcement), was invented in 1824.143. Material for embankment commonlycomes from roadway cuts or designated( borrow areas).144. Causes of automobile accidents canbe categorized into four major groups: thevehicles, the road, the driver, the( pedestrain) .145. Another improvement in drivervisibility is the introduction of theremote-controlled outside ( rearview)mirror.146. Rock nearly always must be drilledand blasted, then loaded with a front-endloader or ( power shovel)into trucks orother hauling units.147. The three forces that can act on astructure are( vertical force), horizontalforce, and those that act upon it with arotating or turning motion.148. Highway pavements are divided intotwo main categories: ( rigid) and flexible.149. Flexible pavements are furtherdivided into three subgroups: high type,( intermediate), and low type.150. The constructing steps of thetransportation system are to plan, design,build, operate and ( maintain).151. The unit price contract is adapted tohighway engineering, because usually it isnot possible to determine exact quantitiesof some items of work ( before)construction is completed.152. The word ‘contract’is derived fromthe Latin ‘contractum’, meaning( drawn)together.153. As a structural material, theenormous advantage of steel is its ( tensilestrength).154. ( Highway transportation)is thedominant transportation mode inpassenger travel.155. The Portland cement concretecommonly used for rigid pavementsconsists of Portland cement, coarseaggregate, ( fine aggregate), water.156. Rigid highway pavement can bedivided into three general types: plainconcrete pavements, simply reinforcedconcrete and ( continuously reinforcedconcrete)pavements.157. The simplest and generally leastcostly form of interchange is the( diamond).158. If distances are great and time is at apremium, ( air)transportation will beselected.159. Signing for freeways should beplanned concurrently with the ( geometry)design.160. Major drainage structures are usuallylarge bridges and multi-span ( culverts).161. The weight of the structure itself isknown as( dead load).162.( Prestressed) concrete is an improvedform of reinforcement.163. A simple contract consists ofan( agreement)entered into by two ormore parties.164. This sum is known as ‘( retention)money’ and serves to insure the employeragainst any defects that may arise in thework.165. Thus,( On-the-job )training can beacquired to translate theory into practiceto the supervisors.166. Large projects ordinarily employ several engineers whose work is coordinated by a( systems engineer). 167. Traffic loads are transferred by the wearing surface to the underlying supporting materials through the interlocking of aggregates, the frictional effect of( granular materials), and cohesion of the fine materials.168. Excavation is the process of loosening and removing earth or rock and transporting it to a fill or to a( waste deposit).169. When planning a structure, an engineer must take into account four factors: dead load,( live load), impact and safety factor.170. The new design standards require( guard)rails and other structures to lessen a vehicle’s impact.171. People select( air transportation)to carry important goods when time is at a premium.172. The benefit-cost ratio method is used for evaluating the( economical)and environmental feasibility of the alternative routes.173. A unique bridge site or a mountain pass also mat become a primary( control point ).174. The radius of a tangent is( infinite), and that of a curve is finite.ing collector-distributor roads can overcome weaving movement of the( cloverleaf) interchange.三．阅读理解题：Passage OneResearch is one of the most important aspects of scientific and engineering practice. A researcher usually works as a member of a team with other scientistsand engineers. He or she is oftenemployed in a laboratory that is financedby government or industry. Areas ofresearch connected with civil engineeringinclude soil mechanics and soilstabilization techniques, and also thedevelopment and the testing of newstructural materials.176. Research is one of ( B. the mostimportant)aspects of scientific andengineering practice.177. A researcher is often employed( C. ina laboratory).178. A researcher usually works as amember of a team with( C. scientists andengineers).179. Which of the following is true?(A.Civil engineering research doesn’t includeonly soil mechanics and soil stabilization,but also the development of new structuralmaterials)Passage TwoThe current tendency is to develop lightermaterials. Aluminum, for example, weighsmuch less than steel but has many of thesame properties. Aluminum beams havealready been used for bridge constructionand for the framework of a few buildings.Attempts are also being made to produceconcrete with more strength and durability,and with a lighter weight. One system thathelps cut concrete weight to some extentuses polymers, which are long chainlikecompounds used in plastics, as part of themixture.180. The current trend of structuralmaterials is( B. to develop lightermaterials).181. Aluminum weighs( A. much less thansteel).182. Aluminum has( C. many of the sameproperties of steel) .183. Which of the following is true?(B.Aluminum beams can be used for not onlybridge construction but also theframework of a few buildings)Passage ThreeSteel and concrete also complement eachother in another way: they have almost thesame rate of contraction and expansion.They therefore can work together insituations where both compression andtension are factors. Steel rods areembedded in concrete to make reinforcedconcrete in concrete beams or structureswhere tension will develop. Concrete andsteel also form such a strong bonds—theforce that unites them—that the steelcannot slip within the concrete. Stillanother advantage is that steel does notrust in concrete. Acid corrodes steel,whereas concrete has an alkaline chemicalreaction, the opposite of acid.184. Steel and concrete have( C. almostthe same rate of contraction andexpansion).185. Reinforced concrete is( A. steel rodswhich are embedded in concrete beams).186. Which of the following is true?(C.steel does not rust in concrete)187. Concrete has( B. an alkalinechemical reaction, the opposite of acid).Passage FourThe employer or promoter of civilengineering works normally determinesthe conditions of contract, which definethe obligations and performances by someform of competitive tendering and anycontractor who submits a successfultender and subsequently enters into acontract is deemed in law to havevoluntarily accepted the conditions of contract adopted by the promoter.The obligations that a contractor accepts when he submits a tender are determined by the form of the invitation to tender. In most cases the tender may be withdrawn at any time until it has been accepted and may, even then, be withdrawn if the acceptance is stated by the promoter to be ‘subject to formal contract’ as is often the case.188. The conditions of contract are normally determined by( C. the promoter). 189. This conditions define the obligations and performances to which (C. the contractor) will be subject.190. The obligations that( C. the contractor)accepts when he submits a tender are determined by the form of the invitation to the tender.191. In most cases the tender may be withdrawn at any time until( B. it has been accepted) .Passage FiveMaterials are usually described as ‘rock’, ‘loose rock’, or ‘common’, with ‘common’signifying all material not otherwise classified. Rock, sometimes called ‘solid rock’, nearly always must be drilled and blasted, then loaded with a front-end loader or power shovel into trucks or other hauling units. Blasted rock may be moved or drifted for short distances by means of a bulldozer, which is, in effect, a huge tractor-mounted blade. Loose rock often is dug with loaders or shovels without any previous blasting. 192. According to the passage, which material signifying all material not otherwise classified.( B. common)193. Which of the following is NOT true?( B. rock, is sometimes called ‘looserock’)194. According to the passage, which ofthe following is true?( B. loose rock isoften dug without any previous blasting)195. Loose rock often is dug with( B.loaders or shovels)without any previousblasting.Passage SixIn the university, mathematics, physics,and chemistry are heavily emphasizedthroughout the engineering curriculum,but particularly in the first two or threeyears. Mathematics is very important inall branches of engineering, so it is greatlystressed. Today, mathematics includescourses in statistics, which deals withgathering, classifying, and usingnumerical data, or pieces of information.An important aspect of statisticalmathematics is probability, which dealswith what may happen when there aredifferent factors, or variables, that canchange the results of a problem. Beforethe construction of a bridge is undertaken,for example, a statistical study is made ofthe amount of traffic the bridge will beexpected to handle. In the design of thebridge, variables such as water pressureon the foundations, impact, the effects ofdifferent wind forces, and many otherfactors must be considered.196. Mathematics is very important in allbranches of engineering so( A. it is greatlystressed).197. Statistics deals with( B. gathering,classifying and using pieces ofinformation).198. An important aspect of statisticalmathematics is( A. probability) .199. Which is the main meaning of thepassage?( B. mathematics is veryimportant in all branches of engineering)Passage SevenCivil engineering projects are almostunique; that is, each has its own problemsand design features. Therefore, carefulstudy is given to each project even beforedesign work begins. The study includes asurvey both of topography and subsoilfeatures of the proposed site. It alsoincludes a consideration of possiblealternatives, such as a concrete gravitydam or an earth-fill embankment dam.The economic factors involved in each ofthe possible alternatives must also beweighed. Today, a study usually includes aconsideration of the environmental impactof the project. Many engineers, usuallyworking as a team that includes surveyors,specialists in soil mechanics, and expertsin design and construction, are involved inmaking these feasibility studies.200. Civil engineering projects are ( A.almost always distinctive)201. Each project( C. must be studiedcarefully)before design work begins202. The study, which must consider notonly structural features but also economicfactors and possible alternatives or otherchoices, is called( B. feasibility study) .203. Which of the following is true?(A.today civil engineering project needconsider the environmental impact of theproject )Passage EightClearing the site precedes all grading andmost other construction operations. Siteclearing in rural areas may sometimesmerely require that glass, shrubs, andother plants or crops be removed.However, it sometimes can involve removing trees and tree stumps and disposing of the debris. The accepted procedure is to remove practically all vegetable matter from the original ground and from fill material, since, if allowed to remain; it may decay and leave voids that result in settlement. Selective clearing in adjoining areas may at times be required. 204. According to the passage, which is the main topic?( B. site clearing)205. According to the passage, ( B. crops) is NOT be removed in rural areas.206. If all vegetable remained,( C. it may decay and leave voids).207. Sometimes it is required clearing( A. adjacent areas) .Passage NineVehicular safety design usually centers on protecting the driver and his passengers in case an accident occurs due to some other failure in the highway system. Examples of this type of design are safety belts and shoulder harnesses, safer door latches, non-shattering windshields, and energy absorbing steering columns. Improvements are made constantly in the parts of a vehicle which are obvious to the driver. These parts include windshield wipers, headlamps, brakes, steering suspension, and the exhaust system. The introduction of front and rear directional signals contributes greatly to motor vehicle safety. Stop lights, backup lights, and four-way emergency flashers also aid in vehicle safety. Four-way emergency flashers have become standard equipment for vehicles.208. According to the passage, which is NOT mentioned?( C. brake pedal)209. According to the passage, which becomes standard equipment forvehicles?( A. four-way emergencyflashers)210. Which of the following is true?( B.Vehicular safety design usually centers onprotecting the passengers and the personswho drives the vehicles)211. According to the passage, whichdesign is described?( A. vehicular safety)Passage TenIn the 1930s engineers found that superiorembankments could be constructed byspreading the material in relatively thinlayers and compacting it at moisturecontent close to optimum. Theimprovement resulted largely becausegreater density was obtained, whichresulted in higher “strength”in the soilmass and in decreased settlement andrutting. Layered construction alsoproduced greater uniformity in thematerial itself and in its density andmoisture content. This was beneficialsince any subsequent consolidation orswelling would be relatively uniform.212. In the 1930s engineers foundembankments could be constructed by( B.compacting it at a moisture content andspreading the material in relatively thinlayers)213. According to the passage, whichcause higher “strength”?( A. greaterdensity was obtained)214. Which of the following is true?( C.layered construction produced greaterunanimity in its density and moisturecontent)215. Which of the following words is theclosest meaning of ‘optimum’?( B. best)Passage ElevenAltogether, three forces can act on astructure: vertical—those that act up ordown; horizontal—those that act sideway;and those that act upon it with a rotatingor turning motion. Forces that act at anangle are combination of horizontal andvertical forces. Since the structuresdesigned by civil engineers are intendedto be stationary or stable, these forcesmust be kept in balance. The verticalforces, for example, must be equal to eachother. If a beam supports a load above, thebeam itself must have sufficient strengthto counterbalance that weight. Thehorizontal forces must also equal eachother so that there is not too much thrusteither to the right or to the left. And forcesthat might pull the structure around mustbe countered with forces that pull in theopposite direction.216. Horizontal forces( B. act sideways).217. Forces acting at an angle arecombination of( A. horizontal and verticalforces).218. The horizontal forces must equaleach other so that( C. there is not toomuch thrust either to the right or to theleft).219. Which of the following is true?( B.three forces acting on a structure must bekept in balance)Passage TwelveWe all enter into contracts almost everyday for the supply of goods, transportationand similar service, and in all theseinstances we are quite willing to pay forthe services we receive. Our needs inthese cases are comparatively simple andwe do not need to enter onto lengthy orcomplicated negotiations and no writtencontract is normally executed.。
T.W. ANDERSON (1971). The Statistical Analysis of Time Series. Series in Probability and Ma
425 BibliographyH.A KAIKE(1974).Markovian representation of stochastic processes and its application to the analysis of autoregressive moving average processes.Annals Institute Statistical Mathematics,vol.26,pp.363-387. B.D.O.A NDERSON and J.B.M OORE(1979).Optimal rmation and System Sciences Series, Prentice Hall,Englewood Cliffs,NJ.T.W.A NDERSON(1971).The Statistical Analysis of Time Series.Series in Probability and Mathematical Statistics,Wiley,New York.R.A NDRE-O BRECHT(1988).A new statistical approach for the automatic segmentation of continuous speech signals.IEEE Trans.Acoustics,Speech,Signal Processing,vol.ASSP-36,no1,pp.29-40.R.A NDRE-O BRECHT(1990).Reconnaissance automatique de parole`a partir de segments acoustiques et de mod`e les de Markov cach´e s.Proc.Journ´e es Etude de la Parole,Montr´e al,May1990(in French).R.A NDRE-O BRECHT and H.Y.S U(1988).Three acoustic labellings for phoneme based continuous speech recognition.Proc.Speech’88,Edinburgh,UK,pp.943-950.U.A PPEL and A.VON B RANDT(1983).Adaptive sequential segmentation of piecewise stationary time rmation Sciences,vol.29,no1,pp.27-56.L.A.A ROIAN and H.L EVENE(1950).The effectiveness of quality control procedures.Jal American Statis-tical Association,vol.45,pp.520-529.K.J.A STR¨OM and B.W ITTENMARK(1984).Computer Controlled Systems:Theory and rma-tion and System Sciences Series,Prentice Hall,Englewood Cliffs,NJ.M.B AGSHAW and R.A.J OHNSON(1975a).The effect of serial correlation on the performance of CUSUM tests-Part II.Technometrics,vol.17,no1,pp.73-80.M.B AGSHAW and R.A.J OHNSON(1975b).The influence of reference values and estimated variance on the ARL of CUSUM tests.Jal Royal Statistical Society,vol.37(B),no3,pp.413-420.M.B AGSHAW and R.A.J OHNSON(1977).Sequential procedures for detecting parameter changes in a time-series model.Jal American Statistical Association,vol.72,no359,pp.593-597.R.K.B ANSAL and P.P APANTONI-K AZAKOS(1986).An algorithm for detecting a change in a stochastic process.IEEE rmation Theory,vol.IT-32,no2,pp.227-235.G.A.B ARNARD(1959).Control charts and stochastic processes.Jal Royal Statistical Society,vol.B.21, pp.239-271.A.E.B ASHARINOV andB.S.F LEISHMAN(1962).Methods of the statistical sequential analysis and their radiotechnical applications.Sovetskoe Radio,Moscow(in Russian).M.B ASSEVILLE(1978).D´e viations par rapport au maximum:formules d’arrˆe t et martingales associ´e es. Compte-rendus du S´e minaire de Probabilit´e s,Universit´e de Rennes I.M.B ASSEVILLE(1981).Edge detection using sequential methods for change in level-Part II:Sequential detection of change in mean.IEEE Trans.Acoustics,Speech,Signal Processing,vol.ASSP-29,no1,pp.32-50.426B IBLIOGRAPHY M.B ASSEVILLE(1982).A survey of statistical failure detection techniques.In Contribution`a la D´e tectionS´e quentielle de Ruptures de Mod`e les Statistiques,Th`e se d’Etat,Universit´e de Rennes I,France(in English). M.B ASSEVILLE(1986).The two-models approach for the on-line detection of changes in AR processes. In Detection of Abrupt Changes in Signals and Dynamical Systems(M.Basseville,A.Benveniste,eds.). Lecture Notes in Control and Information Sciences,LNCIS77,Springer,New York,pp.169-215.M.B ASSEVILLE(1988).Detecting changes in signals and systems-A survey.Automatica,vol.24,pp.309-326.M.B ASSEVILLE(1989).Distance measures for signal processing and pattern recognition.Signal Process-ing,vol.18,pp.349-369.M.B ASSEVILLE and A.B ENVENISTE(1983a).Design and comparative study of some sequential jump detection algorithms for digital signals.IEEE Trans.Acoustics,Speech,Signal Processing,vol.ASSP-31, no3,pp.521-535.M.B ASSEVILLE and A.B ENVENISTE(1983b).Sequential detection of abrupt changes in spectral charac-teristics of digital signals.IEEE rmation Theory,vol.IT-29,no5,pp.709-724.M.B ASSEVILLE and A.B ENVENISTE,eds.(1986).Detection of Abrupt Changes in Signals and Dynamical Systems.Lecture Notes in Control and Information Sciences,LNCIS77,Springer,New York.M.B ASSEVILLE and I.N IKIFOROV(1991).A unified framework for statistical change detection.Proc.30th IEEE Conference on Decision and Control,Brighton,UK.M.B ASSEVILLE,B.E SPIAU and J.G ASNIER(1981).Edge detection using sequential methods for change in level-Part I:A sequential edge detection algorithm.IEEE Trans.Acoustics,Speech,Signal Processing, vol.ASSP-29,no1,pp.24-31.M.B ASSEVILLE, A.B ENVENISTE and G.M OUSTAKIDES(1986).Detection and diagnosis of abrupt changes in modal characteristics of nonstationary digital signals.IEEE rmation Theory,vol.IT-32,no3,pp.412-417.M.B ASSEVILLE,A.B ENVENISTE,G.M OUSTAKIDES and A.R OUG´E E(1987a).Detection and diagnosis of changes in the eigenstructure of nonstationary multivariable systems.Automatica,vol.23,no3,pp.479-489. M.B ASSEVILLE,A.B ENVENISTE,G.M OUSTAKIDES and A.R OUG´E E(1987b).Optimal sensor location for detecting changes in dynamical behavior.IEEE Trans.Automatic Control,vol.AC-32,no12,pp.1067-1075.M.B ASSEVILLE,A.B ENVENISTE,B.G ACH-D EVAUCHELLE,M.G OURSAT,D.B ONNECASE,P.D OREY, M.P REVOSTO and M.O LAGNON(1993).Damage monitoring in vibration mechanics:issues in diagnos-tics and predictive maintenance.Mechanical Systems and Signal Processing,vol.7,no5,pp.401-423.R.V.B EARD(1971).Failure Accommodation in Linear Systems through Self-reorganization.Ph.D.Thesis, Dept.Aeronautics and Astronautics,MIT,Cambridge,MA.A.B ENVENISTE and J.J.F UCHS(1985).Single sample modal identification of a nonstationary stochastic process.IEEE Trans.Automatic Control,vol.AC-30,no1,pp.66-74.A.B ENVENISTE,M.B ASSEVILLE and G.M OUSTAKIDES(1987).The asymptotic local approach to change detection and model validation.IEEE Trans.Automatic Control,vol.AC-32,no7,pp.583-592.A.B ENVENISTE,M.M ETIVIER and P.P RIOURET(1990).Adaptive Algorithms and Stochastic Approxima-tions.Series on Applications of Mathematics,(A.V.Balakrishnan,I.Karatzas,M.Yor,eds.).Springer,New York.A.B ENVENISTE,M.B ASSEVILLE,L.E L G HAOUI,R.N IKOUKHAH and A.S.W ILLSKY(1992).An optimum robust approach to statistical failure detection and identification.IFAC World Conference,Sydney, July1993.B IBLIOGRAPHY427 R.H.B ERK(1973).Some asymptotic aspects of sequential analysis.Annals Statistics,vol.1,no6,pp.1126-1138.R.H.B ERK(1975).Locally most powerful sequential test.Annals Statistics,vol.3,no2,pp.373-381.P.B ILLINGSLEY(1968).Convergence of Probability Measures.Wiley,New York.A.F.B ISSELL(1969).Cusum techniques for quality control.Applied Statistics,vol.18,pp.1-30.M.E.B IVAIKOV(1991).Control of the sample size for recursive estimation of parameters subject to abrupt changes.Automation and Remote Control,no9,pp.96-103.R.E.B LAHUT(1987).Principles and Practice of Information Theory.Addison-Wesley,Reading,MA.I.F.B LAKE and W.C.L INDSEY(1973).Level-crossing problems for random processes.IEEE r-mation Theory,vol.IT-19,no3,pp.295-315.G.B ODENSTEIN and H.M.P RAETORIUS(1977).Feature extraction from the encephalogram by adaptive segmentation.Proc.IEEE,vol.65,pp.642-652.T.B OHLIN(1977).Analysis of EEG signals with changing spectra using a short word Kalman estimator. Mathematical Biosciences,vol.35,pp.221-259.W.B¨OHM and P.H ACKL(1990).Improved bounds for the average run length of control charts based on finite weighted sums.Annals Statistics,vol.18,no4,pp.1895-1899.T.B OJDECKI and J.H OSZA(1984).On a generalized disorder problem.Stochastic Processes and their Applications,vol.18,pp.349-359.L.I.B ORODKIN and V.V.M OTTL’(1976).Algorithm forfinding the jump times of random process equation parameters.Automation and Remote Control,vol.37,no6,Part1,pp.23-32.A.A.B OROVKOV(1984).Theory of Mathematical Statistics-Estimation and Hypotheses Testing,Naouka, Moscow(in Russian).Translated in French under the title Statistique Math´e matique-Estimation et Tests d’Hypoth`e ses,Mir,Paris,1987.G.E.P.B OX and G.M.J ENKINS(1970).Time Series Analysis,Forecasting and Control.Series in Time Series Analysis,Holden-Day,San Francisco.A.VON B RANDT(1983).Detecting and estimating parameters jumps using ladder algorithms and likelihood ratio test.Proc.ICASSP,Boston,MA,pp.1017-1020.A.VON B RANDT(1984).Modellierung von Signalen mit Sprunghaft Ver¨a nderlichem Leistungsspektrum durch Adaptive Segmentierung.Doctor-Engineer Dissertation,M¨u nchen,RFA(in German).S.B RAUN,ed.(1986).Mechanical Signature Analysis-Theory and Applications.Academic Press,London. L.B REIMAN(1968).Probability.Series in Statistics,Addison-Wesley,Reading,MA.G.S.B RITOV and L.A.M IRONOVSKI(1972).Diagnostics of linear systems of automatic regulation.Tekh. Kibernetics,vol.1,pp.76-83.B.E.B RODSKIY and B.S.D ARKHOVSKIY(1992).Nonparametric Methods in Change-point Problems. Kluwer Academic,Boston.L.D.B ROEMELING(1982).Jal Econometrics,vol.19,Special issue on structural change in Econometrics. L.D.B ROEMELING and H.T SURUMI(1987).Econometrics and Structural Change.Dekker,New York. D.B ROOK and D.A.E VANS(1972).An approach to the probability distribution of Cusum run length. Biometrika,vol.59,pp.539-550.J.B RUNET,D.J AUME,M.L ABARR`E RE,A.R AULT and M.V ERG´E(1990).D´e tection et Diagnostic de Pannes.Trait´e des Nouvelles Technologies,S´e rie Diagnostic et Maintenance,Herm`e s,Paris(in French).428B IBLIOGRAPHY S.P.B RUZZONE and M.K AVEH(1984).Information tradeoffs in using the sample autocorrelation function in ARMA parameter estimation.IEEE Trans.Acoustics,Speech,Signal Processing,vol.ASSP-32,no4, pp.701-715.A.K.C AGLAYAN(1980).Necessary and sufficient conditions for detectability of jumps in linear systems. IEEE Trans.Automatic Control,vol.AC-25,no4,pp.833-834.A.K.C AGLAYAN and R.E.L ANCRAFT(1983).Reinitialization issues in fault tolerant systems.Proc.Amer-ican Control Conf.,pp.952-955.A.K.C AGLAYAN,S.M.A LLEN and K.W EHMULLER(1988).Evaluation of a second generation reconfigu-ration strategy for aircraftflight control systems subjected to actuator failure/surface damage.Proc.National Aerospace and Electronic Conference,Dayton,OH.P.E.C AINES(1988).Linear Stochastic Systems.Series in Probability and Mathematical Statistics,Wiley, New York.M.J.C HEN and J.P.N ORTON(1987).Estimation techniques for tracking rapid parameter changes.Intern. Jal Control,vol.45,no4,pp.1387-1398.W.K.C HIU(1974).The economic design of cusum charts for controlling normal mean.Applied Statistics, vol.23,no3,pp.420-433.E.Y.C HOW(1980).A Failure Detection System Design Methodology.Ph.D.Thesis,M.I.T.,L.I.D.S.,Cam-bridge,MA.E.Y.C HOW and A.S.W ILLSKY(1984).Analytical redundancy and the design of robust failure detection systems.IEEE Trans.Automatic Control,vol.AC-29,no3,pp.689-691.Y.S.C HOW,H.R OBBINS and D.S IEGMUND(1971).Great Expectations:The Theory of Optimal Stop-ping.Houghton-Mifflin,Boston.R.N.C LARK,D.C.F OSTH and V.M.W ALTON(1975).Detection of instrument malfunctions in control systems.IEEE Trans.Aerospace Electronic Systems,vol.AES-11,pp.465-473.A.C OHEN(1987).Biomedical Signal Processing-vol.1:Time and Frequency Domain Analysis;vol.2: Compression and Automatic Recognition.CRC Press,Boca Raton,FL.J.C ORGE and F.P UECH(1986).Analyse du rythme cardiaque foetal par des m´e thodes de d´e tection de ruptures.Proc.7th INRIA Int.Conf.Analysis and optimization of Systems.Antibes,FR(in French).D.R.C OX and D.V.H INKLEY(1986).Theoretical Statistics.Chapman and Hall,New York.D.R.C OX and H.D.M ILLER(1965).The Theory of Stochastic Processes.Wiley,New York.S.V.C ROWDER(1987).A simple method for studying run-length distributions of exponentially weighted moving average charts.Technometrics,vol.29,no4,pp.401-407.H.C S¨ORG¨O and L.H ORV´ATH(1988).Nonparametric methods for change point problems.In Handbook of Statistics(P.R.Krishnaiah,C.R.Rao,eds.),vol.7,Elsevier,New York,pp.403-425.R.B.D AVIES(1973).Asymptotic inference in stationary gaussian time series.Advances Applied Probability, vol.5,no3,pp.469-497.J.C.D ECKERT,M.N.D ESAI,J.J.D EYST and A.S.W ILLSKY(1977).F-8DFBW sensor failure identification using analytical redundancy.IEEE Trans.Automatic Control,vol.AC-22,no5,pp.795-803.M.H.D E G ROOT(1970).Optimal Statistical Decisions.Series in Probability and Statistics,McGraw-Hill, New York.J.D ESHAYES and D.P ICARD(1979).Tests de ruptures dans un mod`e pte-Rendus de l’Acad´e mie des Sciences,vol.288,Ser.A,pp.563-566(in French).B IBLIOGRAPHY429 J.D ESHAYES and D.P ICARD(1983).Ruptures de Mod`e les en Statistique.Th`e ses d’Etat,Universit´e deParis-Sud,Orsay,France(in French).J.D ESHAYES and D.P ICARD(1986).Off-line statistical analysis of change-point models using non para-metric and likelihood methods.In Detection of Abrupt Changes in Signals and Dynamical Systems(M. Basseville,A.Benveniste,eds.).Lecture Notes in Control and Information Sciences,LNCIS77,Springer, New York,pp.103-168.B.D EVAUCHELLE-G ACH(1991).Diagnostic M´e canique des Fatigues sur les Structures Soumises`a des Vibrations en Ambiance de Travail.Th`e se de l’Universit´e Paris IX Dauphine(in French).B.D EVAUCHELLE-G ACH,M.B ASSEVILLE and A.B ENVENISTE(1991).Diagnosing mechanical changes in vibrating systems.Proc.SAFEPROCESS’91,Baden-Baden,FRG,pp.85-89.R.D I F RANCESCO(1990).Real-time speech segmentation using pitch and convexity jump models:applica-tion to variable rate speech coding.IEEE Trans.Acoustics,Speech,Signal Processing,vol.ASSP-38,no5, pp.741-748.X.D ING and P.M.F RANK(1990).Fault detection via factorization approach.Systems and Control Letters, vol.14,pp.431-436.J.L.D OOB(1953).Stochastic Processes.Wiley,New York.V.D RAGALIN(1988).Asymptotic solutions in detecting a change in distribution under an unknown param-eter.Statistical Problems of Control,Issue83,Vilnius,pp.45-52.B.D UBUISSON(1990).Diagnostic et Reconnaissance des Formes.Trait´e des Nouvelles Technologies,S´e rie Diagnostic et Maintenance,Herm`e s,Paris(in French).A.J.D UNCAN(1986).Quality Control and Industrial Statistics,5th edition.Richard D.Irwin,Inc.,Home-wood,IL.J.D URBIN(1971).Boundary-crossing probabilities for the Brownian motion and Poisson processes and techniques for computing the power of the Kolmogorov-Smirnov test.Jal Applied Probability,vol.8,pp.431-453.J.D URBIN(1985).Thefirst passage density of the crossing of a continuous Gaussian process to a general boundary.Jal Applied Probability,vol.22,no1,pp.99-122.A.E MAMI-N AEINI,M.M.A KHTER and S.M.R OCK(1988).Effect of model uncertainty on failure detec-tion:the threshold selector.IEEE Trans.Automatic Control,vol.AC-33,no12,pp.1106-1115.J.D.E SARY,F.P ROSCHAN and D.W.W ALKUP(1967).Association of random variables with applications. Annals Mathematical Statistics,vol.38,pp.1466-1474.W.D.E WAN and K.W.K EMP(1960).Sampling inspection of continuous processes with no autocorrelation between successive results.Biometrika,vol.47,pp.263-280.G.F AVIER and A.S MOLDERS(1984).Adaptive smoother-predictors for tracking maneuvering targets.Proc. 23rd Conf.Decision and Control,Las Vegas,NV,pp.831-836.W.F ELLER(1966).An Introduction to Probability Theory and Its Applications,vol.2.Series in Probability and Mathematical Statistics,Wiley,New York.R.A.F ISHER(1925).Theory of statistical estimation.Proc.Cambridge Philosophical Society,vol.22, pp.700-725.M.F ISHMAN(1988).Optimization of the algorithm for the detection of a disorder,based on the statistic of exponential smoothing.In Statistical Problems of Control,Issue83,Vilnius,pp.146-151.R.F LETCHER(1980).Practical Methods of Optimization,2volumes.Wiley,New York.P.M.F RANK(1990).Fault diagnosis in dynamic systems using analytical and knowledge based redundancy -A survey and new results.Automatica,vol.26,pp.459-474.430B IBLIOGRAPHY P.M.F RANK(1991).Enhancement of robustness in observer-based fault detection.Proc.SAFEPRO-CESS’91,Baden-Baden,FRG,pp.275-287.P.M.F RANK and J.W¨UNNENBERG(1989).Robust fault diagnosis using unknown input observer schemes. In Fault Diagnosis in Dynamic Systems-Theory and Application(R.Patton,P.Frank,R.Clark,eds.). International Series in Systems and Control Engineering,Prentice Hall International,London,UK,pp.47-98.K.F UKUNAGA(1990).Introduction to Statistical Pattern Recognition,2d ed.Academic Press,New York. S.I.G ASS(1958).Linear Programming:Methods and Applications.McGraw Hill,New York.W.G E and C.Z.F ANG(1989).Extended robust observation approach for failure isolation.Int.Jal Control, vol.49,no5,pp.1537-1553.W.G ERSCH(1986).Two applications of parametric time series modeling methods.In Mechanical Signature Analysis-Theory and Applications(S.Braun,ed.),chap.10.Academic Press,London.J.J.G ERTLER(1988).Survey of model-based failure detection and isolation in complex plants.IEEE Control Systems Magazine,vol.8,no6,pp.3-11.J.J.G ERTLER(1991).Analytical redundancy methods in fault detection and isolation.Proc.SAFEPRO-CESS’91,Baden-Baden,FRG,pp.9-22.B.K.G HOSH(1970).Sequential Tests of Statistical Hypotheses.Addison-Wesley,Cambridge,MA.I.N.G IBRA(1975).Recent developments in control charts techniques.Jal Quality Technology,vol.7, pp.183-192.J.P.G ILMORE and R.A.M C K ERN(1972).A redundant strapdown inertial reference unit(SIRU).Jal Space-craft,vol.9,pp.39-47.M.A.G IRSHICK and H.R UBIN(1952).A Bayes approach to a quality control model.Annals Mathematical Statistics,vol.23,pp.114-125.A.L.G OEL and S.M.W U(1971).Determination of the ARL and a contour nomogram for CUSUM charts to control normal mean.Technometrics,vol.13,no2,pp.221-230.P.L.G OLDSMITH and H.W HITFIELD(1961).Average run lengths in cumulative chart quality control schemes.Technometrics,vol.3,pp.11-20.G.C.G OODWIN and K.S.S IN(1984).Adaptive Filtering,Prediction and rmation and System Sciences Series,Prentice Hall,Englewood Cliffs,NJ.R.M.G RAY and L.D.D AVISSON(1986).Random Processes:a Mathematical Approach for Engineers. Information and System Sciences Series,Prentice Hall,Englewood Cliffs,NJ.C.G UEGUEN and L.L.S CHARF(1980).Exact maximum likelihood identification for ARMA models:a signal processing perspective.Proc.1st EUSIPCO,Lausanne.D.E.G USTAFSON, A.S.W ILLSKY,J.Y.W ANG,M.C.L ANCASTER and J.H.T RIEBWASSER(1978). ECG/VCG rhythm diagnosis using statistical signal analysis.Part I:Identification of persistent rhythms. Part II:Identification of transient rhythms.IEEE Trans.Biomedical Engineering,vol.BME-25,pp.344-353 and353-361.F.G USTAFSSON(1991).Optimal segmentation of linear regression parameters.Proc.IFAC/IFORS Symp. Identification and System Parameter Estimation,Budapest,pp.225-229.T.H¨AGGLUND(1983).New Estimation Techniques for Adaptive Control.Ph.D.Thesis,Lund Institute of Technology,Lund,Sweden.T.H¨AGGLUND(1984).Adaptive control of systems subject to large parameter changes.Proc.IFAC9th World Congress,Budapest.B IBLIOGRAPHY431 P.H ALL and C.C.H EYDE(1980).Martingale Limit Theory and its Application.Probability and Mathemat-ical Statistics,a Series of Monographs and Textbooks,Academic Press,New York.W.J.H ALL,R.A.W IJSMAN and J.K.G HOSH(1965).The relationship between sufficiency and invariance with applications in sequential analysis.Ann.Math.Statist.,vol.36,pp.576-614.E.J.H ANNAN and M.D EISTLER(1988).The Statistical Theory of Linear Systems.Series in Probability and Mathematical Statistics,Wiley,New York.J.D.H EALY(1987).A note on multivariate CuSum procedures.Technometrics,vol.29,pp.402-412.D.M.H IMMELBLAU(1970).Process Analysis by Statistical Methods.Wiley,New York.D.M.H IMMELBLAU(1978).Fault Detection and Diagnosis in Chemical and Petrochemical Processes. Chemical Engineering Monographs,vol.8,Elsevier,Amsterdam.W.G.S.H INES(1976a).A simple monitor of a system with sudden parameter changes.IEEE r-mation Theory,vol.IT-22,no2,pp.210-216.W.G.S.H INES(1976b).Improving a simple monitor of a system with sudden parameter changes.IEEE rmation Theory,vol.IT-22,no4,pp.496-499.D.V.H INKLEY(1969).Inference about the intersection in two-phase regression.Biometrika,vol.56,no3, pp.495-504.D.V.H INKLEY(1970).Inference about the change point in a sequence of random variables.Biometrika, vol.57,no1,pp.1-17.D.V.H INKLEY(1971).Inference about the change point from cumulative sum-tests.Biometrika,vol.58, no3,pp.509-523.D.V.H INKLEY(1971).Inference in two-phase regression.Jal American Statistical Association,vol.66, no336,pp.736-743.J.R.H UDDLE(1983).Inertial navigation system error-model considerations in Kalmanfiltering applica-tions.In Control and Dynamic Systems(C.T.Leondes,ed.),Academic Press,New York,pp.293-339.J.S.H UNTER(1986).The exponentially weighted moving average.Jal Quality Technology,vol.18,pp.203-210.I.A.I BRAGIMOV and R.Z.K HASMINSKII(1981).Statistical Estimation-Asymptotic Theory.Applications of Mathematics Series,vol.16.Springer,New York.R.I SERMANN(1984).Process fault detection based on modeling and estimation methods-A survey.Auto-matica,vol.20,pp.387-404.N.I SHII,A.I WATA and N.S UZUMURA(1979).Segmentation of nonstationary time series.Int.Jal Systems Sciences,vol.10,pp.883-894.J.E.J ACKSON and R.A.B RADLEY(1961).Sequential and tests.Annals Mathematical Statistics, vol.32,pp.1063-1077.B.J AMES,K.L.J AMES and D.S IEGMUND(1988).Conditional boundary crossing probabilities with appli-cations to change-point problems.Annals Probability,vol.16,pp.825-839.M.K.J EERAGE(1990).Reliability analysis of fault-tolerant IMU architectures with redundant inertial sen-sors.IEEE Trans.Aerospace and Electronic Systems,vol.AES-5,no.7,pp.23-27.N.L.J OHNSON(1961).A simple theoretical approach to cumulative sum control charts.Jal American Sta-tistical Association,vol.56,pp.835-840.N.L.J OHNSON and F.C.L EONE(1962).Cumulative sum control charts:mathematical principles applied to their construction and use.Parts I,II,III.Industrial Quality Control,vol.18,pp.15-21;vol.19,pp.29-36; vol.20,pp.22-28.432B IBLIOGRAPHY R.A.J OHNSON and M.B AGSHAW(1974).The effect of serial correlation on the performance of CUSUM tests-Part I.Technometrics,vol.16,no.1,pp.103-112.H.L.J ONES(1973).Failure Detection in Linear Systems.Ph.D.Thesis,Dept.Aeronautics and Astronautics, MIT,Cambridge,MA.R.H.J ONES,D.H.C ROWELL and L.E.K APUNIAI(1970).Change detection model for serially correlated multivariate data.Biometrics,vol.26,no2,pp.269-280.M.J URGUTIS(1984).Comparison of the statistical properties of the estimates of the change times in an autoregressive process.In Statistical Problems of Control,Issue65,Vilnius,pp.234-243(in Russian).T.K AILATH(1980).Linear rmation and System Sciences Series,Prentice Hall,Englewood Cliffs,NJ.L.V.K ANTOROVICH and V.I.K RILOV(1958).Approximate Methods of Higher Analysis.Interscience,New York.S.K ARLIN and H.M.T AYLOR(1975).A First Course in Stochastic Processes,2d ed.Academic Press,New York.S.K ARLIN and H.M.T AYLOR(1981).A Second Course in Stochastic Processes.Academic Press,New York.D.K AZAKOS and P.P APANTONI-K AZAKOS(1980).Spectral distance measures between gaussian pro-cesses.IEEE Trans.Automatic Control,vol.AC-25,no5,pp.950-959.K.W.K EMP(1958).Formula for calculating the operating characteristic and average sample number of some sequential tests.Jal Royal Statistical Society,vol.B-20,no2,pp.379-386.K.W.K EMP(1961).The average run length of the cumulative sum chart when a V-mask is used.Jal Royal Statistical Society,vol.B-23,pp.149-153.K.W.K EMP(1967a).Formal expressions which can be used for the determination of operating character-istics and average sample number of a simple sequential test.Jal Royal Statistical Society,vol.B-29,no2, pp.248-262.K.W.K EMP(1967b).A simple procedure for determining upper and lower limits for the average sample run length of a cumulative sum scheme.Jal Royal Statistical Society,vol.B-29,no2,pp.263-265.D.P.K ENNEDY(1976).Some martingales related to cumulative sum tests and single server queues.Stochas-tic Processes and Appl.,vol.4,pp.261-269.T.H.K ERR(1980).Statistical analysis of two-ellipsoid overlap test for real time failure detection.IEEE Trans.Automatic Control,vol.AC-25,no4,pp.762-772.T.H.K ERR(1982).False alarm and correct detection probabilities over a time interval for restricted classes of failure detection algorithms.IEEE rmation Theory,vol.IT-24,pp.619-631.T.H.K ERR(1987).Decentralizedfiltering and redundancy management for multisensor navigation.IEEE Trans.Aerospace and Electronic systems,vol.AES-23,pp.83-119.Minor corrections on p.412and p.599 (May and July issues,respectively).R.A.K HAN(1978).Wald’s approximations to the average run length in cusum procedures.Jal Statistical Planning and Inference,vol.2,no1,pp.63-77.R.A.K HAN(1979).Somefirst passage problems related to cusum procedures.Stochastic Processes and Applications,vol.9,no2,pp.207-215.R.A.K HAN(1981).A note on Page’s two-sided cumulative sum procedures.Biometrika,vol.68,no3, pp.717-719.B IBLIOGRAPHY433 V.K IREICHIKOV,V.M ANGUSHEV and I.N IKIFOROV(1990).Investigation and application of CUSUM algorithms to monitoring of sensors.In Statistical Problems of Control,Issue89,Vilnius,pp.124-130(in Russian).G.K ITAGAWA and W.G ERSCH(1985).A smoothness prior time-varying AR coefficient modeling of non-stationary covariance time series.IEEE Trans.Automatic Control,vol.AC-30,no1,pp.48-56.N.K LIGIENE(1980).Probabilities of deviations of the change point estimate in statistical models.In Sta-tistical Problems of Control,Issue83,Vilnius,pp.80-86(in Russian).N.K LIGIENE and L.T ELKSNYS(1983).Methods of detecting instants of change of random process prop-erties.Automation and Remote Control,vol.44,no10,Part II,pp.1241-1283.J.K ORN,S.W.G ULLY and A.S.W ILLSKY(1982).Application of the generalized likelihood ratio algorithm to maneuver detection and estimation.Proc.American Control Conf.,Arlington,V A,pp.792-798.P.R.K RISHNAIAH and B.Q.M IAO(1988).Review about estimation of change points.In Handbook of Statistics(P.R.Krishnaiah,C.R.Rao,eds.),vol.7,Elsevier,New York,pp.375-402.P.K UDVA,N.V ISWANADHAM and A.R AMAKRISHNAN(1980).Observers for linear systems with unknown inputs.IEEE Trans.Automatic Control,vol.AC-25,no1,pp.113-115.S.K ULLBACK(1959).Information Theory and Statistics.Wiley,New York(also Dover,New York,1968). K.K UMAMARU,S.S AGARA and T.S¨ODERSTR¨OM(1989).Some statistical methods for fault diagnosis for dynamical systems.In Fault Diagnosis in Dynamic Systems-Theory and Application(R.Patton,P.Frank,R. Clark,eds.).International Series in Systems and Control Engineering,Prentice Hall International,London, UK,pp.439-476.A.K USHNIR,I.N IKIFOROV and I.S AVIN(1983).Statistical adaptive algorithms for automatic detection of seismic signals-Part I:One-dimensional case.In Earthquake Prediction and the Study of the Earth Structure,Naouka,Moscow(Computational Seismology,vol.15),pp.154-159(in Russian).L.L ADELLI(1990).Diffusion approximation for a pseudo-likelihood test process with application to de-tection of change in stochastic system.Stochastics and Stochastics Reports,vol.32,pp.1-25.T.L.L A¨I(1974).Control charts based on weighted sums.Annals Statistics,vol.2,no1,pp.134-147.T.L.L A¨I(1981).Asymptotic optimality of invariant sequential probability ratio tests.Annals Statistics, vol.9,no2,pp.318-333.D.G.L AINIOTIS(1971).Joint detection,estimation,and system identifirmation and Control, vol.19,pp.75-92.M.R.L EADBETTER,G.L INDGREN and H.R OOTZEN(1983).Extremes and Related Properties of Random Sequences and Processes.Series in Statistics,Springer,New York.L.L E C AM(1960).Locally asymptotically normal families of distributions.Univ.California Publications in Statistics,vol.3,pp.37-98.L.L E C AM(1986).Asymptotic Methods in Statistical Decision Theory.Series in Statistics,Springer,New York.E.L.L EHMANN(1986).Testing Statistical Hypotheses,2d ed.Wiley,New York.J.P.L EHOCZKY(1977).Formulas for stopped diffusion processes with stopping times based on the maxi-mum.Annals Probability,vol.5,no4,pp.601-607.H.R.L ERCHE(1980).Boundary Crossing of Brownian Motion.Lecture Notes in Statistics,vol.40,Springer, New York.L.L JUNG(1987).System Identification-Theory for the rmation and System Sciences Series, Prentice Hall,Englewood Cliffs,NJ.。
3GPP TS 36.331 V13.2.0 (2016-06)
3GPP TS 36.331 V13.2.0 (2016-06)Technical Specification3rd Generation Partnership Project;Technical Specification Group Radio Access Network;Evolved Universal Terrestrial Radio Access (E-UTRA);Radio Resource Control (RRC);Protocol specification(Release 13)The present document has been developed within the 3rd Generation Partnership Project (3GPP TM) and may be further elaborated for the purposes of 3GPP. The present document has not been subject to any approval process by the 3GPP Organizational Partners and shall not be implemented.This Specification is provided for future development work within 3GPP only. The Organizational Partners accept no liability for any use of this Specification. Specifications and reports for implementation of the 3GPP TM system should be obtained via the 3GPP Organizational Partners' Publications Offices.KeywordsUMTS, radio3GPPPostal address3GPP support office address650 Route des Lucioles - Sophia AntipolisValbonne - FRANCETel.: +33 4 92 94 42 00 Fax: +33 4 93 65 47 16InternetCopyright NotificationNo part may be reproduced except as authorized by written permission.The copyright and the foregoing restriction extend to reproduction in all media.© 2016, 3GPP Organizational Partners (ARIB, ATIS, CCSA, ETSI, TSDSI, TTA, TTC).All rights reserved.UMTS™ is a Trade Mark of ETSI registered for the benefit of its members3GPP™ is a Trade Mark of ETSI registered for the benefit of its Members and of the 3GPP Organizational PartnersLTE™ is a Trade Mark of ETSI currently being registered for the benefit of its Members and of the 3GPP Organizational Partners GSM® and the GSM logo are registered and owned by the GSM AssociationBluetooth® is a Trade Mark of the Bluetooth SIG registered for the benefit of its membersContentsForeword (18)1Scope (19)2References (19)3Definitions, symbols and abbreviations (22)3.1Definitions (22)3.2Abbreviations (24)4General (27)4.1Introduction (27)4.2Architecture (28)4.2.1UE states and state transitions including inter RAT (28)4.2.2Signalling radio bearers (29)4.3Services (30)4.3.1Services provided to upper layers (30)4.3.2Services expected from lower layers (30)4.4Functions (30)5Procedures (32)5.1General (32)5.1.1Introduction (32)5.1.2General requirements (32)5.2System information (33)5.2.1Introduction (33)5.2.1.1General (33)5.2.1.2Scheduling (34)5.2.1.2a Scheduling for NB-IoT (34)5.2.1.3System information validity and notification of changes (35)5.2.1.4Indication of ETWS notification (36)5.2.1.5Indication of CMAS notification (37)5.2.1.6Notification of EAB parameters change (37)5.2.1.7Access Barring parameters change in NB-IoT (37)5.2.2System information acquisition (38)5.2.2.1General (38)5.2.2.2Initiation (38)5.2.2.3System information required by the UE (38)5.2.2.4System information acquisition by the UE (39)5.2.2.5Essential system information missing (42)5.2.2.6Actions upon reception of the MasterInformationBlock message (42)5.2.2.7Actions upon reception of the SystemInformationBlockType1 message (42)5.2.2.8Actions upon reception of SystemInformation messages (44)5.2.2.9Actions upon reception of SystemInformationBlockType2 (44)5.2.2.10Actions upon reception of SystemInformationBlockType3 (45)5.2.2.11Actions upon reception of SystemInformationBlockType4 (45)5.2.2.12Actions upon reception of SystemInformationBlockType5 (45)5.2.2.13Actions upon reception of SystemInformationBlockType6 (45)5.2.2.14Actions upon reception of SystemInformationBlockType7 (45)5.2.2.15Actions upon reception of SystemInformationBlockType8 (45)5.2.2.16Actions upon reception of SystemInformationBlockType9 (46)5.2.2.17Actions upon reception of SystemInformationBlockType10 (46)5.2.2.18Actions upon reception of SystemInformationBlockType11 (46)5.2.2.19Actions upon reception of SystemInformationBlockType12 (47)5.2.2.20Actions upon reception of SystemInformationBlockType13 (48)5.2.2.21Actions upon reception of SystemInformationBlockType14 (48)5.2.2.22Actions upon reception of SystemInformationBlockType15 (48)5.2.2.23Actions upon reception of SystemInformationBlockType16 (48)5.2.2.24Actions upon reception of SystemInformationBlockType17 (48)5.2.2.25Actions upon reception of SystemInformationBlockType18 (48)5.2.2.26Actions upon reception of SystemInformationBlockType19 (49)5.2.3Acquisition of an SI message (49)5.2.3a Acquisition of an SI message by BL UE or UE in CE or a NB-IoT UE (50)5.3Connection control (50)5.3.1Introduction (50)5.3.1.1RRC connection control (50)5.3.1.2Security (52)5.3.1.2a RN security (53)5.3.1.3Connected mode mobility (53)5.3.1.4Connection control in NB-IoT (54)5.3.2Paging (55)5.3.2.1General (55)5.3.2.2Initiation (55)5.3.2.3Reception of the Paging message by the UE (55)5.3.3RRC connection establishment (56)5.3.3.1General (56)5.3.3.1a Conditions for establishing RRC Connection for sidelink communication/ discovery (58)5.3.3.2Initiation (59)5.3.3.3Actions related to transmission of RRCConnectionRequest message (63)5.3.3.3a Actions related to transmission of RRCConnectionResumeRequest message (64)5.3.3.4Reception of the RRCConnectionSetup by the UE (64)5.3.3.4a Reception of the RRCConnectionResume by the UE (66)5.3.3.5Cell re-selection while T300, T302, T303, T305, T306, or T308 is running (68)5.3.3.6T300 expiry (68)5.3.3.7T302, T303, T305, T306, or T308 expiry or stop (69)5.3.3.8Reception of the RRCConnectionReject by the UE (70)5.3.3.9Abortion of RRC connection establishment (71)5.3.3.10Handling of SSAC related parameters (71)5.3.3.11Access barring check (72)5.3.3.12EAB check (73)5.3.3.13Access barring check for ACDC (73)5.3.3.14Access Barring check for NB-IoT (74)5.3.4Initial security activation (75)5.3.4.1General (75)5.3.4.2Initiation (76)5.3.4.3Reception of the SecurityModeCommand by the UE (76)5.3.5RRC connection reconfiguration (77)5.3.5.1General (77)5.3.5.2Initiation (77)5.3.5.3Reception of an RRCConnectionReconfiguration not including the mobilityControlInfo by theUE (77)5.3.5.4Reception of an RRCConnectionReconfiguration including the mobilityControlInfo by the UE(handover) (79)5.3.5.5Reconfiguration failure (83)5.3.5.6T304 expiry (handover failure) (83)5.3.5.7Void (84)5.3.5.7a T307 expiry (SCG change failure) (84)5.3.5.8Radio Configuration involving full configuration option (84)5.3.6Counter check (86)5.3.6.1General (86)5.3.6.2Initiation (86)5.3.6.3Reception of the CounterCheck message by the UE (86)5.3.7RRC connection re-establishment (87)5.3.7.1General (87)5.3.7.2Initiation (87)5.3.7.3Actions following cell selection while T311 is running (88)5.3.7.4Actions related to transmission of RRCConnectionReestablishmentRequest message (89)5.3.7.5Reception of the RRCConnectionReestablishment by the UE (89)5.3.7.6T311 expiry (91)5.3.7.7T301 expiry or selected cell no longer suitable (91)5.3.7.8Reception of RRCConnectionReestablishmentReject by the UE (91)5.3.8RRC connection release (92)5.3.8.1General (92)5.3.8.2Initiation (92)5.3.8.3Reception of the RRCConnectionRelease by the UE (92)5.3.8.4T320 expiry (93)5.3.9RRC connection release requested by upper layers (93)5.3.9.1General (93)5.3.9.2Initiation (93)5.3.10Radio resource configuration (93)5.3.10.0General (93)5.3.10.1SRB addition/ modification (94)5.3.10.2DRB release (95)5.3.10.3DRB addition/ modification (95)5.3.10.3a1DC specific DRB addition or reconfiguration (96)5.3.10.3a2LWA specific DRB addition or reconfiguration (98)5.3.10.3a3LWIP specific DRB addition or reconfiguration (98)5.3.10.3a SCell release (99)5.3.10.3b SCell addition/ modification (99)5.3.10.3c PSCell addition or modification (99)5.3.10.4MAC main reconfiguration (99)5.3.10.5Semi-persistent scheduling reconfiguration (100)5.3.10.6Physical channel reconfiguration (100)5.3.10.7Radio Link Failure Timers and Constants reconfiguration (101)5.3.10.8Time domain measurement resource restriction for serving cell (101)5.3.10.9Other configuration (102)5.3.10.10SCG reconfiguration (103)5.3.10.11SCG dedicated resource configuration (104)5.3.10.12Reconfiguration SCG or split DRB by drb-ToAddModList (105)5.3.10.13Neighbour cell information reconfiguration (105)5.3.10.14Void (105)5.3.10.15Sidelink dedicated configuration (105)5.3.10.16T370 expiry (106)5.3.11Radio link failure related actions (107)5.3.11.1Detection of physical layer problems in RRC_CONNECTED (107)5.3.11.2Recovery of physical layer problems (107)5.3.11.3Detection of radio link failure (107)5.3.12UE actions upon leaving RRC_CONNECTED (109)5.3.13UE actions upon PUCCH/ SRS release request (110)5.3.14Proximity indication (110)5.3.14.1General (110)5.3.14.2Initiation (111)5.3.14.3Actions related to transmission of ProximityIndication message (111)5.3.15Void (111)5.4Inter-RAT mobility (111)5.4.1Introduction (111)5.4.2Handover to E-UTRA (112)5.4.2.1General (112)5.4.2.2Initiation (112)5.4.2.3Reception of the RRCConnectionReconfiguration by the UE (112)5.4.2.4Reconfiguration failure (114)5.4.2.5T304 expiry (handover to E-UTRA failure) (114)5.4.3Mobility from E-UTRA (114)5.4.3.1General (114)5.4.3.2Initiation (115)5.4.3.3Reception of the MobilityFromEUTRACommand by the UE (115)5.4.3.4Successful completion of the mobility from E-UTRA (116)5.4.3.5Mobility from E-UTRA failure (117)5.4.4Handover from E-UTRA preparation request (CDMA2000) (117)5.4.4.1General (117)5.4.4.2Initiation (118)5.4.4.3Reception of the HandoverFromEUTRAPreparationRequest by the UE (118)5.4.5UL handover preparation transfer (CDMA2000) (118)5.4.5.1General (118)5.4.5.2Initiation (118)5.4.5.3Actions related to transmission of the ULHandoverPreparationTransfer message (119)5.4.5.4Failure to deliver the ULHandoverPreparationTransfer message (119)5.4.6Inter-RAT cell change order to E-UTRAN (119)5.4.6.1General (119)5.4.6.2Initiation (119)5.4.6.3UE fails to complete an inter-RAT cell change order (119)5.5Measurements (120)5.5.1Introduction (120)5.5.2Measurement configuration (121)5.5.2.1General (121)5.5.2.2Measurement identity removal (122)5.5.2.2a Measurement identity autonomous removal (122)5.5.2.3Measurement identity addition/ modification (123)5.5.2.4Measurement object removal (124)5.5.2.5Measurement object addition/ modification (124)5.5.2.6Reporting configuration removal (126)5.5.2.7Reporting configuration addition/ modification (127)5.5.2.8Quantity configuration (127)5.5.2.9Measurement gap configuration (127)5.5.2.10Discovery signals measurement timing configuration (128)5.5.2.11RSSI measurement timing configuration (128)5.5.3Performing measurements (128)5.5.3.1General (128)5.5.3.2Layer 3 filtering (131)5.5.4Measurement report triggering (131)5.5.4.1General (131)5.5.4.2Event A1 (Serving becomes better than threshold) (135)5.5.4.3Event A2 (Serving becomes worse than threshold) (136)5.5.4.4Event A3 (Neighbour becomes offset better than PCell/ PSCell) (136)5.5.4.5Event A4 (Neighbour becomes better than threshold) (137)5.5.4.6Event A5 (PCell/ PSCell becomes worse than threshold1 and neighbour becomes better thanthreshold2) (138)5.5.4.6a Event A6 (Neighbour becomes offset better than SCell) (139)5.5.4.7Event B1 (Inter RAT neighbour becomes better than threshold) (139)5.5.4.8Event B2 (PCell becomes worse than threshold1 and inter RAT neighbour becomes better thanthreshold2) (140)5.5.4.9Event C1 (CSI-RS resource becomes better than threshold) (141)5.5.4.10Event C2 (CSI-RS resource becomes offset better than reference CSI-RS resource) (141)5.5.4.11Event W1 (WLAN becomes better than a threshold) (142)5.5.4.12Event W2 (All WLAN inside WLAN mobility set becomes worse than threshold1 and a WLANoutside WLAN mobility set becomes better than threshold2) (142)5.5.4.13Event W3 (All WLAN inside WLAN mobility set becomes worse than a threshold) (143)5.5.5Measurement reporting (144)5.5.6Measurement related actions (148)5.5.6.1Actions upon handover and re-establishment (148)5.5.6.2Speed dependant scaling of measurement related parameters (149)5.5.7Inter-frequency RSTD measurement indication (149)5.5.7.1General (149)5.5.7.2Initiation (150)5.5.7.3Actions related to transmission of InterFreqRSTDMeasurementIndication message (150)5.6Other (150)5.6.0General (150)5.6.1DL information transfer (151)5.6.1.1General (151)5.6.1.2Initiation (151)5.6.1.3Reception of the DLInformationTransfer by the UE (151)5.6.2UL information transfer (151)5.6.2.1General (151)5.6.2.2Initiation (151)5.6.2.3Actions related to transmission of ULInformationTransfer message (152)5.6.2.4Failure to deliver ULInformationTransfer message (152)5.6.3UE capability transfer (152)5.6.3.1General (152)5.6.3.2Initiation (153)5.6.3.3Reception of the UECapabilityEnquiry by the UE (153)5.6.4CSFB to 1x Parameter transfer (157)5.6.4.1General (157)5.6.4.2Initiation (157)5.6.4.3Actions related to transmission of CSFBParametersRequestCDMA2000 message (157)5.6.4.4Reception of the CSFBParametersResponseCDMA2000 message (157)5.6.5UE Information (158)5.6.5.1General (158)5.6.5.2Initiation (158)5.6.5.3Reception of the UEInformationRequest message (158)5.6.6 Logged Measurement Configuration (159)5.6.6.1General (159)5.6.6.2Initiation (160)5.6.6.3Reception of the LoggedMeasurementConfiguration by the UE (160)5.6.6.4T330 expiry (160)5.6.7 Release of Logged Measurement Configuration (160)5.6.7.1General (160)5.6.7.2Initiation (160)5.6.8 Measurements logging (161)5.6.8.1General (161)5.6.8.2Initiation (161)5.6.9In-device coexistence indication (163)5.6.9.1General (163)5.6.9.2Initiation (164)5.6.9.3Actions related to transmission of InDeviceCoexIndication message (164)5.6.10UE Assistance Information (165)5.6.10.1General (165)5.6.10.2Initiation (166)5.6.10.3Actions related to transmission of UEAssistanceInformation message (166)5.6.11 Mobility history information (166)5.6.11.1General (166)5.6.11.2Initiation (166)5.6.12RAN-assisted WLAN interworking (167)5.6.12.1General (167)5.6.12.2Dedicated WLAN offload configuration (167)5.6.12.3WLAN offload RAN evaluation (167)5.6.12.4T350 expiry or stop (167)5.6.12.5Cell selection/ re-selection while T350 is running (168)5.6.13SCG failure information (168)5.6.13.1General (168)5.6.13.2Initiation (168)5.6.13.3Actions related to transmission of SCGFailureInformation message (168)5.6.14LTE-WLAN Aggregation (169)5.6.14.1Introduction (169)5.6.14.2Reception of LWA configuration (169)5.6.14.3Release of LWA configuration (170)5.6.15WLAN connection management (170)5.6.15.1Introduction (170)5.6.15.2WLAN connection status reporting (170)5.6.15.2.1General (170)5.6.15.2.2Initiation (171)5.6.15.2.3Actions related to transmission of WLANConnectionStatusReport message (171)5.6.15.3T351 Expiry (WLAN connection attempt timeout) (171)5.6.15.4WLAN status monitoring (171)5.6.16RAN controlled LTE-WLAN interworking (172)5.6.16.1General (172)5.6.16.2WLAN traffic steering command (172)5.6.17LTE-WLAN aggregation with IPsec tunnel (173)5.6.17.1General (173)5.7Generic error handling (174)5.7.1General (174)5.7.2ASN.1 violation or encoding error (174)5.7.3Field set to a not comprehended value (174)5.7.4Mandatory field missing (174)5.7.5Not comprehended field (176)5.8MBMS (176)5.8.1Introduction (176)5.8.1.1General (176)5.8.1.2Scheduling (176)5.8.1.3MCCH information validity and notification of changes (176)5.8.2MCCH information acquisition (178)5.8.2.1General (178)5.8.2.2Initiation (178)5.8.2.3MCCH information acquisition by the UE (178)5.8.2.4Actions upon reception of the MBSFNAreaConfiguration message (178)5.8.2.5Actions upon reception of the MBMSCountingRequest message (179)5.8.3MBMS PTM radio bearer configuration (179)5.8.3.1General (179)5.8.3.2Initiation (179)5.8.3.3MRB establishment (179)5.8.3.4MRB release (179)5.8.4MBMS Counting Procedure (179)5.8.4.1General (179)5.8.4.2Initiation (180)5.8.4.3Reception of the MBMSCountingRequest message by the UE (180)5.8.5MBMS interest indication (181)5.8.5.1General (181)5.8.5.2Initiation (181)5.8.5.3Determine MBMS frequencies of interest (182)5.8.5.4Actions related to transmission of MBMSInterestIndication message (183)5.8a SC-PTM (183)5.8a.1Introduction (183)5.8a.1.1General (183)5.8a.1.2SC-MCCH scheduling (183)5.8a.1.3SC-MCCH information validity and notification of changes (183)5.8a.1.4Procedures (184)5.8a.2SC-MCCH information acquisition (184)5.8a.2.1General (184)5.8a.2.2Initiation (184)5.8a.2.3SC-MCCH information acquisition by the UE (184)5.8a.2.4Actions upon reception of the SCPTMConfiguration message (185)5.8a.3SC-PTM radio bearer configuration (185)5.8a.3.1General (185)5.8a.3.2Initiation (185)5.8a.3.3SC-MRB establishment (185)5.8a.3.4SC-MRB release (185)5.9RN procedures (186)5.9.1RN reconfiguration (186)5.9.1.1General (186)5.9.1.2Initiation (186)5.9.1.3Reception of the RNReconfiguration by the RN (186)5.10Sidelink (186)5.10.1Introduction (186)5.10.1a Conditions for sidelink communication operation (187)5.10.2Sidelink UE information (188)5.10.2.1General (188)5.10.2.2Initiation (189)5.10.2.3Actions related to transmission of SidelinkUEInformation message (193)5.10.3Sidelink communication monitoring (195)5.10.6Sidelink discovery announcement (198)5.10.6a Sidelink discovery announcement pool selection (201)5.10.6b Sidelink discovery announcement reference carrier selection (201)5.10.7Sidelink synchronisation information transmission (202)5.10.7.1General (202)5.10.7.2Initiation (203)5.10.7.3Transmission of SLSS (204)5.10.7.4Transmission of MasterInformationBlock-SL message (205)5.10.7.5Void (206)5.10.8Sidelink synchronisation reference (206)5.10.8.1General (206)5.10.8.2Selection and reselection of synchronisation reference UE (SyncRef UE) (206)5.10.9Sidelink common control information (207)5.10.9.1General (207)5.10.9.2Actions related to reception of MasterInformationBlock-SL message (207)5.10.10Sidelink relay UE operation (207)5.10.10.1General (207)5.10.10.2AS-conditions for relay related sidelink communication transmission by sidelink relay UE (207)5.10.10.3AS-conditions for relay PS related sidelink discovery transmission by sidelink relay UE (208)5.10.10.4Sidelink relay UE threshold conditions (208)5.10.11Sidelink remote UE operation (208)5.10.11.1General (208)5.10.11.2AS-conditions for relay related sidelink communication transmission by sidelink remote UE (208)5.10.11.3AS-conditions for relay PS related sidelink discovery transmission by sidelink remote UE (209)5.10.11.4Selection and reselection of sidelink relay UE (209)5.10.11.5Sidelink remote UE threshold conditions (210)6Protocol data units, formats and parameters (tabular & ASN.1) (210)6.1General (210)6.2RRC messages (212)6.2.1General message structure (212)–EUTRA-RRC-Definitions (212)–BCCH-BCH-Message (212)–BCCH-DL-SCH-Message (212)–BCCH-DL-SCH-Message-BR (213)–MCCH-Message (213)–PCCH-Message (213)–DL-CCCH-Message (214)–DL-DCCH-Message (214)–UL-CCCH-Message (214)–UL-DCCH-Message (215)–SC-MCCH-Message (215)6.2.2Message definitions (216)–CounterCheck (216)–CounterCheckResponse (217)–CSFBParametersRequestCDMA2000 (217)–CSFBParametersResponseCDMA2000 (218)–DLInformationTransfer (218)–HandoverFromEUTRAPreparationRequest (CDMA2000) (219)–InDeviceCoexIndication (220)–InterFreqRSTDMeasurementIndication (222)–LoggedMeasurementConfiguration (223)–MasterInformationBlock (225)–MBMSCountingRequest (226)–MBMSCountingResponse (226)–MBMSInterestIndication (227)–MBSFNAreaConfiguration (228)–MeasurementReport (228)–MobilityFromEUTRACommand (229)–Paging (232)–ProximityIndication (233)–RNReconfiguration (234)–RNReconfigurationComplete (234)–RRCConnectionReconfiguration (235)–RRCConnectionReconfigurationComplete (240)–RRCConnectionReestablishment (241)–RRCConnectionReestablishmentComplete (241)–RRCConnectionReestablishmentReject (242)–RRCConnectionReestablishmentRequest (243)–RRCConnectionReject (243)–RRCConnectionRelease (244)–RRCConnectionResume (248)–RRCConnectionResumeComplete (249)–RRCConnectionResumeRequest (250)–RRCConnectionRequest (250)–RRCConnectionSetup (251)–RRCConnectionSetupComplete (252)–SCGFailureInformation (253)–SCPTMConfiguration (254)–SecurityModeCommand (255)–SecurityModeComplete (255)–SecurityModeFailure (256)–SidelinkUEInformation (256)–SystemInformation (258)–SystemInformationBlockType1 (259)–UEAssistanceInformation (264)–UECapabilityEnquiry (265)–UECapabilityInformation (266)–UEInformationRequest (267)–UEInformationResponse (267)–ULHandoverPreparationTransfer (CDMA2000) (273)–ULInformationTransfer (274)–WLANConnectionStatusReport (274)6.3RRC information elements (275)6.3.1System information blocks (275)–SystemInformationBlockType2 (275)–SystemInformationBlockType3 (279)–SystemInformationBlockType4 (282)–SystemInformationBlockType5 (283)–SystemInformationBlockType6 (287)–SystemInformationBlockType7 (289)–SystemInformationBlockType8 (290)–SystemInformationBlockType9 (295)–SystemInformationBlockType10 (295)–SystemInformationBlockType11 (296)–SystemInformationBlockType12 (297)–SystemInformationBlockType13 (297)–SystemInformationBlockType14 (298)–SystemInformationBlockType15 (298)–SystemInformationBlockType16 (299)–SystemInformationBlockType17 (300)–SystemInformationBlockType18 (301)–SystemInformationBlockType19 (301)–SystemInformationBlockType20 (304)6.3.2Radio resource control information elements (304)–AntennaInfo (304)–AntennaInfoUL (306)–CQI-ReportConfig (307)–CQI-ReportPeriodicProcExtId (314)–CrossCarrierSchedulingConfig (314)–CSI-IM-Config (315)–CSI-IM-ConfigId (315)–CSI-RS-Config (317)–CSI-RS-ConfigEMIMO (318)–CSI-RS-ConfigNZP (319)–CSI-RS-ConfigNZPId (320)–CSI-RS-ConfigZP (321)–CSI-RS-ConfigZPId (321)–DMRS-Config (321)–DRB-Identity (322)–EPDCCH-Config (322)–EIMTA-MainConfig (324)–LogicalChannelConfig (325)–LWA-Configuration (326)–LWIP-Configuration (326)–RCLWI-Configuration (327)–MAC-MainConfig (327)–P-C-AndCBSR (332)–PDCCH-ConfigSCell (333)–PDCP-Config (334)–PDSCH-Config (337)–PDSCH-RE-MappingQCL-ConfigId (339)–PHICH-Config (339)–PhysicalConfigDedicated (339)–P-Max (344)–PRACH-Config (344)–PresenceAntennaPort1 (346)–PUCCH-Config (347)–PUSCH-Config (351)–RACH-ConfigCommon (355)–RACH-ConfigDedicated (357)–RadioResourceConfigCommon (358)–RadioResourceConfigDedicated (362)–RLC-Config (367)–RLF-TimersAndConstants (369)–RN-SubframeConfig (370)–SchedulingRequestConfig (371)–SoundingRS-UL-Config (372)–SPS-Config (375)–TDD-Config (376)–TimeAlignmentTimer (377)–TPC-PDCCH-Config (377)–TunnelConfigLWIP (378)–UplinkPowerControl (379)–WLAN-Id-List (382)–WLAN-MobilityConfig (382)6.3.3Security control information elements (382)–NextHopChainingCount (382)–SecurityAlgorithmConfig (383)–ShortMAC-I (383)6.3.4Mobility control information elements (383)–AdditionalSpectrumEmission (383)–ARFCN-ValueCDMA2000 (383)–ARFCN-ValueEUTRA (384)–ARFCN-ValueGERAN (384)–ARFCN-ValueUTRA (384)–BandclassCDMA2000 (384)–BandIndicatorGERAN (385)–CarrierFreqCDMA2000 (385)–CarrierFreqGERAN (385)–CellIndexList (387)–CellReselectionPriority (387)–CellSelectionInfoCE (387)–CellReselectionSubPriority (388)–CSFB-RegistrationParam1XRTT (388)–CellGlobalIdEUTRA (389)–CellGlobalIdUTRA (389)–CellGlobalIdGERAN (390)–CellGlobalIdCDMA2000 (390)–CellSelectionInfoNFreq (391)–CSG-Identity (391)–FreqBandIndicator (391)–MobilityControlInfo (391)–MobilityParametersCDMA2000 (1xRTT) (393)–MobilityStateParameters (394)–MultiBandInfoList (394)–NS-PmaxList (394)–PhysCellId (395)–PhysCellIdRange (395)–PhysCellIdRangeUTRA-FDDList (395)–PhysCellIdCDMA2000 (396)–PhysCellIdGERAN (396)–PhysCellIdUTRA-FDD (396)–PhysCellIdUTRA-TDD (396)–PLMN-Identity (397)–PLMN-IdentityList3 (397)–PreRegistrationInfoHRPD (397)–Q-QualMin (398)–Q-RxLevMin (398)–Q-OffsetRange (398)–Q-OffsetRangeInterRAT (399)–ReselectionThreshold (399)–ReselectionThresholdQ (399)–SCellIndex (399)–ServCellIndex (400)–SpeedStateScaleFactors (400)–SystemInfoListGERAN (400)–SystemTimeInfoCDMA2000 (401)–TrackingAreaCode (401)–T-Reselection (402)–T-ReselectionEUTRA-CE (402)6.3.5Measurement information elements (402)–AllowedMeasBandwidth (402)–CSI-RSRP-Range (402)–Hysteresis (402)–LocationInfo (403)–MBSFN-RSRQ-Range (403)–MeasConfig (404)–MeasDS-Config (405)–MeasGapConfig (406)–MeasId (407)–MeasIdToAddModList (407)–MeasObjectCDMA2000 (408)–MeasObjectEUTRA (408)–MeasObjectGERAN (412)–MeasObjectId (412)–MeasObjectToAddModList (412)–MeasObjectUTRA (413)–ReportConfigEUTRA (422)–ReportConfigId (425)–ReportConfigInterRAT (425)–ReportConfigToAddModList (428)–ReportInterval (429)–RSRP-Range (429)–RSRQ-Range (430)–RSRQ-Type (430)–RS-SINR-Range (430)–RSSI-Range-r13 (431)–TimeToTrigger (431)–UL-DelayConfig (431)–WLAN-CarrierInfo (431)–WLAN-RSSI-Range (432)–WLAN-Status (432)6.3.6Other information elements (433)–AbsoluteTimeInfo (433)–AreaConfiguration (433)–C-RNTI (433)–DedicatedInfoCDMA2000 (434)–DedicatedInfoNAS (434)–FilterCoefficient (434)–LoggingDuration (434)–LoggingInterval (435)–MeasSubframePattern (435)–MMEC (435)–NeighCellConfig (435)–OtherConfig (436)–RAND-CDMA2000 (1xRTT) (437)–RAT-Type (437)–ResumeIdentity (437)–RRC-TransactionIdentifier (438)–S-TMSI (438)–TraceReference (438)–UE-CapabilityRAT-ContainerList (438)–UE-EUTRA-Capability (439)–UE-RadioPagingInfo (469)–UE-TimersAndConstants (469)–VisitedCellInfoList (470)–WLAN-OffloadConfig (470)6.3.7MBMS information elements (472)–MBMS-NotificationConfig (472)–MBMS-ServiceList (473)–MBSFN-AreaId (473)–MBSFN-AreaInfoList (473)–MBSFN-SubframeConfig (474)–PMCH-InfoList (475)6.3.7a SC-PTM information elements (476)–SC-MTCH-InfoList (476)–SCPTM-NeighbourCellList (478)6.3.8Sidelink information elements (478)–SL-CommConfig (478)–SL-CommResourcePool (479)–SL-CP-Len (480)–SL-DiscConfig (481)–SL-DiscResourcePool (483)–SL-DiscTxPowerInfo (485)–SL-GapConfig (485)。
A posteriori error bounds for the zeros of polynomials
A posteriori error bounds for the zeros o/" polynomials*
By W. BORSCH-SUPAN 1. Introduction In most cases, the computed solution of a certain problem in numerical analysis is only an approximation to the true solution of the problem, since there are errors originating from discretization or truncation and from rounding. This is true also for problems of the algebraic kind like a system of linear equations, the matrix eigenvalue problem, the factorization of polynomials, etc. Because of rounding errors, it is even true for the results obtained b y some direct method, e.g. the inversion of a matrix b y the Gaul3 elimination technique. In general, sufficiently close bounds for the errors of those approximations can be obtained a posteriori only, i.e. the formulae for close bounds in general contain final and/or intermediate results of the computation. An important subclass of such estimates does not use intermediate results. This means that the estimates are applicable independently of the method employed for solving the problem. This is particularly true for the estimates which use the difference between the respective right hand sides and left hand sides of a system of equations which is to be solved. These estimates m a y be called "defect estimates" or "residual estimates" analogous to the German terminus "Defektabsch~itzungen". This paper is concerned with estimates of this type. As already pointed out b y WILKINSON I9J, the considerable amount of work involved in obtaining close error bounds makes it desirable to improve the given approximate results at the same time, and to give bounds for the improved approximations, too. In the case of matrix inversion, such improvements can be achieved b y iteration according to SCHULZ ~7], and the respective estimates follow from certain inequalities of the second kind for suitable matrix norms. For the matrix eigenvalue-vector problem, WILKINSON [9] has given corresponding results. The situation is very similar in the case of a polynomial equation, which is the subject of this paper. A common feature to these three problems is, that the residuals, which are the basic data used for the estimation, are very sensitive to rounding errors. Therefore an analysis of these errors must be given, too. 2. The simplest error bounds for single roots In this section, we restrict our considerations to polynomials with well isolated single complex roots. Multiple and clustering roots will be discussed in section 4. * Part of this work was done during the author's stay with the National Bureau of Standards, Washington, D.C.
新SAT评分详解及样题
* Combined score of two raters, each scoring on a 1– 4 scale 1-4
SAT 1. Composite Score 2 2. SAT raw score 3 3. SAT Test Score Evidence-Based Reading and Writing raw score 4. SAT Studies OG 1—15 5. SAT Subscore 7 Cross-section Score 3 Section Score 400—1600
3.
)
25+15min 49
:35min 44
2-12 25min
2-8 50min
History Questions 1-5 are based on the following passage.
This passage is adapted from a speech delivered by Congresswoman Barbara Jordan of Texas on July 25, 1974, as a member of the Judiciary Committee of the United States House of Representatives. In the passage, Jordan discusses how and when a United States president may be impeached, or charged with serious offenses, while in office. Jordan’s speech was delivered in the context of impeachment hearings against then president Richard M. Nixon.
考虑撬棒的双馈型风场集电线速断保护
电气传动2022年第52卷第7期摘要：汇流集电线故障是并网双馈型风电场最为常见的故障之一。
由于撬棒系统中卸流电阻的影响，双馈风机在汇流集电线发生不对称故障后，其正、负序阻抗将会根据转差率的不同产生不同的特性，从而影响传统自适应电流速断保护的选择性，造成风电机组的大规模误切除，降低双馈型风电场内部的可靠性与并网稳定性。
鉴于此，为了改善双馈型风电场的继电保护性能，基于考虑撬棒动作后的双馈风机阻抗特性，提出了一种适用于35kV 汇流集电线的自适应电流速断保护，并对其具体的整定计算进行了详细的分析。
基于PSCAD 的仿真模型验证了所提保护方法与整定计算的正确性与有效性。
关键词：汇流线路短路故障；双馈型风场；电流速断保护；整定计算；阻抗特性中图分类号：TM28文献标识码：ADOI ：10.19457/j.1001-2095.dqcd22194Instantaneous Protection for the Collector Lines of Doubly -fed Wind Farm Considering CrowbarLI Xuhui 1，XIE Baihuang 2，HUANG Xiaoyong 1，XU Yan 3（1.Shangluo Power Supply Bureau ，State Grid Shaanxi Electric Power Company ，Shangluo726000，Shaanxi ，China ；2.Control Center ，Shaanxi Power Supply Bureau ，Xi 'an 710048，Shaanxi ，China ；3.Department of Electric Power Engineering ，North ChinaElectric Power University ，Baoding 071003，Hebei ，China ）Abstract:The collector line short-circuit fault is one of the most common faults for the grid-connected doubly-fed wind farms.Because the introduction of the crowbar discharge resistor ，after the unsymmetrical faults of the collector line ，the positive and negative sequence impedances of DFIG will produce the different characteristics according to the different slip rate.Then the selectivity of the traditional adaptive instantaneous overcurret protection is affected and a good deal of DFIGs may be cutted off incorrectly ，the internal reliability and grid-connection stability of the doubly-fed wind farm are reduced.In view of the above situation ，an improved adaptive instantaneous overcurret protection （IAIOP ）based on the impedance characteristics of the crowbar circuit was proposed for the 35kV collector line to improve the protection performance of the doubly-fed wind farm.Meanwhile ，the setting calculation of the proposed IAIOP was also analyzed in detail.The correctness and effectiveness of the proposed protection method and setting calculation were validated by the simulation model in the PSCAD.Key words:short-circuit fault of the collector line ；doubly-fed wind farm ；instantaneous overcurret protection ；setting calculation ；impedance characteristics基金项目：国家自然科学基金项目（51307059）作者简介：李旭辉（1977—），女，本科，副高级工程师，Email ：通讯作者：徐岩（1976—），男，博士，副教授，Email ：考虑撬棒的双馈型风场集电线速断保护李旭辉1，谢百煌2，黄晓勇1，徐岩3（1.国网陕西省电力公司商洛供电公司，陕西商洛726000；2.国网陕西省电力公司调控中心，陕西西安710048；3.华北电力大学电力工程系，河北保定071003）作为目前技术最为成熟、经济效益最高的可再生能源[1-2]，风电系统的应用得到了能源可持续发展研究领域的广泛关注。
2015-FAST--RAIDShield_Characterizing, Monitoring, and Proactively Protecting Against Disk Failures
RAIDShield:Characterizing,Monitoring,and Proactively ProtectingAgainst Disk FailuresAo Ma1,Fred Douglis1,Guanlin Lu1,Darren Sawyer1,Surendar Chandra2,Windsor Hsu21EMC Corporation,2Datrium,Inc.AbstractModern storage systems orchestrate a group of disks to achieve their performance and reliability goals.Even though such systems are designed to withstand the fail-ure of individual disks,failure of multiple disks poses a unique set of challenges.We empirically investigate disk failure data from a large number of production systems, specifically focusing on the impact of disk failures on RAID storage systems.Our data covers about one million SATA disks from6disk models for periods up to 5years.We show how observed disk failures weaken the protection provided by RAID.The count of reallocated sectors correlates strongly with impending failures. With thesefindings we designed RAIDS HIELD, which consists of two components.First,we have built and evaluated an active defense mechanism that moni-tors the health of each disk and replaces those that are predicted to fail imminently.This proactive protection has been incorporated into our product and is observed to eliminate88%of triple disk errors,which are80%of all RAID failures.Second,we have designed and simulated a method of using the joint failure probability to quantify and predict how likely a RAID group is to face multi-ple simultaneous disk failures,which can identify disks that collectively represent a risk of failure even when no individual disk isflagged in isolation.Wefind in sim-ulation that RAID-level analysis can effectively identify most vulnerable RAID-6systems,improving the cover-age to98%of triple errors.1IntroductionStorage systems have relied for decades on redundancy mechanisms such as RAID to tolerate disk failures,as-suming an ideal world with independent and instanta-neous failures as well as exponential distributions of the time to failure[3,11,18,36].However,some assump-tions no longer hold given the fault model presented by modern disk drives.Schroeder and Gibson[42]analyzed 100,000disks and rejected the hypothesis of the time be-tween disk replacements following an exponential distri-bution.Further,in addition to whole-disk failures that make an entire disk unusable,modern drives can exhibit latent sector errors in which a block or set of blocks be-come inaccessible[6,29].Such sector faults in otherwise working disks further weaken the RAID reconstruction capability.Not only were sector errors previously ig-nored in the early RAID reliability model,these errors may worsen over time due to increasing drive complex-ity[4]and the common use of less reliable disks in stor-age systems[6,17,38].In short,RAID protection is no longer enough;however,given its prevalence in the stor-age industry,a mechanism to shield RAID systems from unreliable disks would have a wide audience.System designers have realized the new threats caused by these disk faults and built additional mechanisms to improve data reliability.While the original RAID sys-tem would protect against the loss of data from one disk(either an unavailable sector or the failure of the entire disk),the trend has been to use additional re-dundancy to guard against related data loss on mul-tiple disks.For example,some storage arrays incor-porate extra levels of parity,such as RAID-6,which can tolerate two simultaneous whole or partial disk fail-ures[2,12,13,19,22,23];others add redundancy with CPU-intensive erasure coding[14,25].Throughout this paper we focus on“triple-disk failures,”or“triple fail-ures”for short,which refer to any combination of losing related data from three disks simultaneously,due to bad sectors or an entire disk.If a RAID-6system encoun-ters a triple failure it will lose data,but additional layers of redundancy(such as replication)can further protect against catastrophic data loss.Many storage systems apply disk scrubbing to proac-tively detect latent sector errors;i.e.,they read data from disk specifically to check for media errors,rather than be-cause an application has requested the data[28,43].File systems also incorporate techniques such as replication and parity to improve data availability[10,37,41];repli-cation is critical because the failure of a disk group(DG) can be rectified,at high overhead,with a separate replica accessible via a LAN or W AN.Finally,even when pri-mary storage systems are backed up onto separate ded-icated backup systems,those backup systems can them-selves be replicated[27].Unfortunately,improvements to the basic RAID archi-tecture are still based on certain assumptions given the limited understanding of disk fault modes.For example, empirical observations show both the sector error rate and the whole-disk failure rate grow over time[6,42], causing RAID availability to continuously degrade.It is possible for multiple disks in the same RAID DG to fail simultaneously while other working disks have de-veloped a number of latent sector errors[16].Such mul-tiple combined faults can overcome RAID protection and affect data availability.Unfortunately,little data is pub-licly available that quantifies such correlated faults.To address this knowledge gap with respect to storage system reliability,we collected and analyzed disk error logs from EMC Data Domain backup systems.The data cover periods up to60months and include about1mil-lion SATA disks from deployed systems at customer and internal sites.To our knowledge,this is thefirst study of this magnitude to focus on analyzing disk faults(e.g., whole-disk failures and sector errors)that influence data reliability.The logs report when a disk failure is de-tected,at which point a system can automatically initi-ate data recovery onto a spare drive using available data from within that system.They also report larger-scale outages,when too many drives fail simultaneously for data to be accessible.We define a recovery-related inci-dent as a failure that requires the retrieval of data from another system,such as a backup or disk replica.Our analysis reveals that many disks fail at a similar age and the frequency of sector errors keeps increasing on working disks.Ensuring data reliability in the worst case requires adding considerable extra redundancy, making the traditional passive approach of RAID pro-tection unattractive from a cost perspective.By studying numerous types of disk error,we also observe that the accumulation of sector errors contributes to whole-disk failures,causing disk reliability to deteriorate continu-ously.Specifically,a large number of reallocated sectors (RS1)indicates a high probability of imminent whole-disk failure or,at a minimum,a burst of sector errors. With thesefindings we designed RAIDS HIELD,a monitoring mechanism,which proactively identifies and preempts impending failures and vulnerable RAID groups.RAIDS HIELD consists of two components, P LATE+A RMOR.First,we have built and evaluated Pre-dict Loss Accumulating in The Enterprise(P LATE),an active defense mechanism that monitors the health of each disk by tracking the number of reallocated sec-tors,proactively detecting unstable disks and replacing them in advance.P LATE has been deployed in produc-tion systems for nearly a year.Second,we have de-1RS is also sometimes referred to as RAS in disk statistics,but we prefer to avoid the confusion with other uses of RAS in the CS literature.signed and simulated Assure Redundant Media Or Re-place(A RMOR),which uses the joint failure probabil-ity of a DG to quantify the likelihood of multiple si-multaneous disk failures.A RMOR has the potential to identify sets of disks that collectively represent a risk of failure even when no individual disk isflagged in iso-lation.Given this assessment,unstable disks can then be replaced in advance or the redundancy of a DG can be increased;either approach can improve overall RAID availability.Simulation results for P LATE,the single-disk proac-tive protection,show it can capture up to65%of impend-ing whole-disk failures with up to2.5%false alarms.Af-ter incorporating it into our product,wefind its effect on RAID failures is disproportionate:it has been observed to eliminate70%of the recovery-related incidents caused by RAID failures and88%of the RAID failures due to triple disk failures.Its benefits are somewhat limited by the types of errors that it cannot predict:about20%of DG failures are caused by user errors,hardware faults, and other unknown reasons.Simulation results indicate that A RMOR,the cross-disk proactive protection,can ef-fectively identify80%of vulnerable RAID-6systems in a test of5500DGs.Wefind that it can predict most of the triple failures not prevented by P LATE,leading to to-tal coverage of98%of triple failures.The rest of this paper is organized as follows.Wefirst provide background on partial disk failures and describe our storage system architecture,including an overview of RAIDS HIELD(§2).§3presents our study on the rela-tion between whole-disk failure and sector errors,and it characterizes reallocated sectors,which are found to be highly correlated with whole-disk failures.§4describes and evaluates P LATE,demonstrating the substantial re-duction in RAID failures after deploying single-disk pre-dictive replacement.§5describes the design and evalua-tion,via simulation,of A RMOR:using joint probabilities to assess the failure risk to a DG as a whole.§6discusses related work and§7concludes.2Background and MotivationIn this section we define disk partial failures,provid-ing the background to understand our subsequent failure analysis.We then present an overview of our storage sys-tem architecture and describe the two aspects of RAID-S HIELD.2.1Disk FailuresDisks do not fail in a simple fail-stop fashion.Hence, there is no consensus definition of what constitutes a disk failure[5,8,45].The production systems we studied define a whole-disk failure as:•The system loses its connection to the disk,•An operation exceeds the timeout threshold,or •A write operation fails.These criteria serve as the bottom line to replace disks that cannot function properly.However,in addition to whole-disk failures,disk drives can experience vari-ous partial failures while they still otherwise function. Sector-related issues are the major partial failures that endanger data safety[7,31,41].Disk drives therefore provide a variety of proprietary and complicated mecha-nisms to rectify some failures and extend drive lifespans. In this subsection,we briefly describe disk technology, focusing on detection and error handling mechanisms for sector errors;refer elsewhere for more detailed descrip-tions[6,38].Failure detection and recovery mechanisms vary by manufacturer,production model,interface and capacity;the mechanisms introduced here cover com-mon SATA disk internal mechanisms.Sector errors can be categorized into different specific types based on how they are detected,as shown in Fig-ure1.Operations to the disk can be initiated byfile sys-tem read()and write()calls as well as by an internal scan process,which systematically checks sector reliability and accessibility in the background.(These are shown in Figure1in blue,magenta,and green respectively.) Media error:This error occurs when a particular disk sector cannot be read,whether during a normal read or a background disk scan.Any data previously stored in the sector is lost.The disk interface reports the status code upon detecting a sector error,specifying the reason why the read command failed.Pending and Uncorrectable sector:Unstable sectors detected in the background process will be marked as pending sectors,and disk drives can try rectifying these errors through internal protection mechanisms,such as built-in Error Correcting Codes and Refreshment.These techniques rewrite the sector with the data read from that track to recover the faded data.Any sectors that are not successfully recovered will be marked as uncorrectable sectors.Reallocated sector:After a number of unsuccessful re-tries,disk drives automatically re-map a failed write to a spare sector;its logical block address(LBA)remains unchanged.Modern disk drives usually reserve a few thousand spare sectors,which are not initially mapped to particular LBAs.Reallocation only occurs on detected write errors.We also observe that changes to disk technology tend to increase the frequency of sector errors,a major frac-tion of partial disk failures.First,the number of sectors on a disk keeps increasing:while the capacity of individ-ual disks may not be increasing at the rate once predicted by Kryder[33,47],they still increase.Thus,if sector er-rors occur at the current rate,there would be moresector Figure1:Sector error transition.Thisfigure depicts different responses to sector errors.A read(shown in blue)will report a media error if target sector is unreadable.A write (magenta)will attempt to remap a bad sector.An internal scan (green)will try to identify and rectify unstable sectors. errors per disk.Second,the disk capacity increase comes from packing more sectors per track,rather than adding more physical platters.Sectors become increasingly vul-nerable to media scratches and side-track erasures[15].2.2Storage System EnvironmentWe now briefly describe the context of our storage sys-tem with a focus on sector error detection and handling. At a high level,the storage system is composed of three layers,including a typicalfile system,the RAID layer, and the storage layer.Thefile system processes client re-quests by sending read and write operations to the RAID layer.The RAID layer transforms thefile system re-quests into disk logical block requests and passes them to the storage layer,which accesses the physical disks.Our RAID layer adopts the RAID-6algorithm,which can tol-erate two simultaneous failures.In addition to reporting latent sector errors captured in ordinary I/Os,our storage systems scrub all disks pe-riodically as a proactive measure to detect latent sector errors and data corruption errors.Specifically,this scan process checks the accessibility of“live”sectors(those storing data accessible through thefile system),verifies the checksums,and notifies the RAID layer on failures. Sector error handling depends on the type of disk re-quest.A failed write is re-directed to a spare sector through the automatic disk remapping process,without reporting the error to the storage layer.If a read fails,the RAID layer reconstructs data on the inaccessible sector and passes it to the storage layer for rewriting.Writing to the failed sector will trigger the disk internal mapping process.Note that given the process of RAID recon-struction and re-issued write,the failed sector detectedthrough read (media error)will eventually lead to an RS.Therefore,the RS count is actually the number of inac-cessible sectors detected in either reads or writes.Finally,the systems evaluated in this paper are backup systems,which are known to have write-heavy work-loads with fewer random I/Os than primary storage [46];this workload may change the way in which disk faults are detected,as write errors may be relatively more com-mon than read errors.The general conclusions should hold for other types of use.2.3RAIDS HIELD MotivationDespite the expectation that RAID-6systems should be resilient to disk failures,given a large enough popula-tion of DGs there will be errors leading to potential data loss [3].Indeed,our systems encounter RAID-level er-rors,but thankfully these are extremely rare.2These systems usually rely on extra layers of redundancy such as (possibly off-site)replication to guard against catas-trophic failures,but there is a strong incentive to decrease the rate at which RAID failures occur.As we see in §3,disks that are installed together are somewhat likely to fail together,and disks that have par-tial (media)errors will rapidly accumulate errors until they are deemed to have failed completely.Our goal for RAIDS HIELD is to identify and replace failing disks be-fore they completely fail,within reason.In the extreme case,one could use a single disk error as a warning signal and replace any disk as soon as it reported the slightest problem.However,the cost in time and expense would be prohibitive,especially for large-scale installations like cloud providers.With RAIDS HIELD ,we take two tacks in this regard.The first is to use statistical information to discriminate between those disks that are likely to fail soon and those that are not.In the next section we con-sider a number of disk statistics that might be used for this purpose,finding that the reallocated sectors (RS)metric is an excellent predictor of impending failures.We show in §4that after deploying P LATE proactive disk replacement,looking at each disk in isolation,our RAID failures dropped dramatically.Can we do better with A RMOR ,our second tack?We hypothesize that by using the joint failure probability across a DG we can find some additional instances where no single disk is close enough to failure to justify replac-ing it using the criteria for P LATE ,but enough disks are symptomatic that the DG as a whole is in jeopardy.In §5we present the probability analysis and some simulation results to justify this approach.In addition,we specu-late that in some environments,it will be undesirable to2Weare unable to release specific error rates for DGs or disk mod-els.Figure 2:Example of RAIDS HIELD .Four DGs areshown,each with four disks.Green disks are healthy,yellow disks are at risk,and red disks are likely to fail imminently.DG 2and DG 3are at risk of failure.proactively replace every disk that is showing the possi-bility of failure;instead,it may be important to prioritize among DGs and first replace disks in the most vulner-able groups.A single soon-to-fail disk in an otherwise healthy DG is a lower risk than a DG with many disks that have moderate probability of failure.Figure 2provides an example of the difference be-tween P LATE and A RMOR .There are four disk groups;DG 2,with two failing disks,is at high risk,while DG 3has a moderate risk due to the large number of partly-failing disks.With P LATE ,we would replace the red disks,protecting vulnerable DG 2and improving the pro-tection of DG 4,but DG 4is already protected by three healthy disks.With A RMOR ,we replace the two failing disks in DG 2but also recognize the vulnerability of DG 3given the large number of at-risk disks.3Disk Failure AnalysisUnderstanding the nature of whole-disk failures and par-tial failures is essential for improving storage system re-liability and availability.This section presents the results of our analysis of about 1million SATA disks.First,we describe how we collected the disk data studied in this work.Second,we present our observations of the new disk failure modes (e.g.,simultaneous disk failures and sector errors)which endanger RAID availability.Third,we analyze the correlation between these two failure modes.Finally,we analyze characteristics and proper-ties of reallocated sectors,the specific sector error type that is found to predict drive failures.P e r c e n t a g emonthsmonthsmonthsP e r c e n t a g emonthsmonthsmonthsFigure 3:Distribution of lifetimes of failed drives.These graphs show that many disks fail at a similar age.Note that thenumber of buckets,i.e.total age since deployment,and time length of each bucket varies by drive.Disk Population First Log Length Model (Thousands)Deployment (Months)A-13406/200860A-216511/200860B-110006/200848C-19310/201036C-225312/201036D-138409/201121Table 1:Disk population.Population,earliest deploy-ment date and log length of disk models used in this study.3.1Data CollectionOur storage system has a built-in mechanism to log sys-tem status,which can optionally send important events back to a central repository each day [46].These mes-sages record a variety of system events including disk errors and failures.The data studied here are collected from these reports over a period of 5years starting in June,2008.Similar to previous work [6],we anonymize disk in-formation to make it possible to compare across disks from a single manufacturer but not across disk families.We denote each disk drive model as family -capacity .Family is a single letter representing the disk family and capacity is a single number representing the disk’s partic-ular capacity.Although capacities are anonymized as a single number,relative sizes within a family are ordered by the number representing the capacity.That is,A-2and C-2are larger than A-1and C-1respectively.Our entire sample of 1million disks includes 6disk models,each of which has a population of at least 30,000.They have been shipped in our storage systems since June,2008,giving us a sufficient observation win-dow to study various errors over the full lifespans of many drives.Details of the drives studied are presented in Table 1.Note that the recorded period of each disk model varies:the studied data range from 60-month logs of A-1and A-2down to 21months for D-1.3.2New Disk Failure ModesWe observe two new disk failure modes that are not pre-dicted by the early RAID reliability model and degrade RAID reliability and availability.Drives fail at similar ages:We analyze all failed drives and categorize them into different buckets based on their lifetime.Figure 3shows that a large fraction of failed drives are found at a similar age.For example,63%of A-1failed drives,66%of A-2failed drives and 64%of B-1failed drives are found in their fourth year.This fail-ure peak is also observed in the second year of the C-2model,with 68%of failed drives found in this period.Given a large population of drives,some drives will fail not only in the same month but occasionally the same week or day,resulting in vulnerable systems.If a third error (a defective sector or a failed drive)should also oc-cur before drives can be replaced and data reconstructed,the DG will be unavailable.The lifetime distributions of C-1and D-1failed drives are comparatively uniform.However,these drives are510152025F r a c t i o n o f E r r o r D i s k s (%)369121518212427303336months A-1A-2B-1C-1C-2D-1Figure 4:Percentage of disks developing sector er-rors.As disks age,the number with at least one errorincreases,and the rate of increase is higher the older the disk is.Note that D-1has only a 21-month record.A v e r a g e R e a l l o c a t e d S e c t o r C o u n tDisk ModelFigure 5:Error counts year over year.Among diskswith sector errors,for each model the number of errors in-creased significantly in the second year.relatively young compared to the drives with long obser-vation intervals,so it is difficult to draw specific conclu-sions from this uniformity.We note a degree of “infant mortality”with these drives,with peaks of failures in the first three months.Sector errors exacerbate risk:Figure 4presents the fraction of disks affected by sector errors as a function of the disk age.Disks from all models show sector er-rors by the time they have been in use for 2–3years,but some have significant errors much earlier.In addition,the rate at which errors appear increases with the age of the disks:for example,about 5%of A-2disks get sector errors in the first 30months,but it only takes an addi-tional 6months for 10%more to develop sector errors.Similar trends can be observed with A-1,B-1,and C-2.To demonstrate the rate of error increase,we select 1000disks randomly from each disk model,which de-veloped at least one sector in a one-month observation window.We collect the count of their sector errors one year later.Figure 5shows the average number of sector errors in the first and second years.For all drives with at least one sector error,the number of sector errors for the second year increases considerably,ranging from 25%for the C-2model to about 300%for A-2.These new disk failure modes reveal that the tradi-tional RAID mechanism has become inadequate.The observation that many disks fail at a similar age means RAID systems face a higher risk of multiple whole-disk failures than anticipated.The increasing frequency of sector errors in working disks means RAID systems face a correspondingly higher risk of reconstruction failures:a disk that has not completely failed may be unable to provide specific sectors needed for the reconstruction.The disk technology trends introduced in §2.1exacerbate these risks.3.3Correlating Full and Partial ErrorsSince both whole-disk failures and sector errors affect data availability,exploring how they are correlated helps us to understand the challenges of RAID reliability.Here we introduce the statistical methodology used to analyze the data,then we evaluate the correlation between whole-disk failures and sector errors.3.3.1Statistical MethodsOur objective is to compare the sector errors in working disks and failed ones,and to use a measure to reflect their discrimination.We use quantile distributions to quantita-tively evaluate the correlation degree between disk fail-ures and sector errors.Specifically,we collect the num-ber of sector errors on working and failed disks,summa-rizing each data set value using deciles of the cumulative distribution (i.e.,we divide the sorted data set into ten equal-sized subsets;we normally display only the first nine deciles to avoid the skew of outliers).Such quan-tiles are more robust than other statistical techniques,such as mean and cumulative distribution function,to outliers and noise in depicting the value distribution and have been used to analyze performance crises in data centers [9].3.3.2Identifying CorrelationAs introduced in §2.1,sector errors can be categorized into specific types based on how they are detected.For example,a sector error detected in a read is regarded as a media error while a sector error captured in a write is counted as an RS.Those error counts can be collected through the disk SMART interface [1]and are included in our logs.Figures 6-7compare the deciles of disk errors built on1002003004005006007008009001000R e a l l o c a t e d S e c t o r C o u n t123456789Deciles05285317133251174390200000001failed disk deciles working disk decilesA-120040060080010001200140016001800200022002400R e a l l o c a t e d S e c t o r C o u n t123456789Deciles2238718732752281212422025000012629failed disk deciles working disk decilesA-220040060080010001200140016001800200022002400R e a l l o c a t e d S e c t o r C o u n t123456789Deciles5491132203385698181342203200001213failed disk deciles working disk decilesB-130060090012001500R e a l l o c a t e d S e c t o r C o u n t123456789Deciles0000015226697400000000failed disk deciles working disk decilesC-130060090012001500R e a l l o c a t e d S e c t o r C o u n t123456789Deciles 139881432213245047661364000000000failed disk deciles working disk decilesC-2100200300400500R e a l l o c a t e d S e c t o r C o u n t123456789Deciles000182671160393000000000failed disk deciles working disk decilesD-1Figure 6:Reallocated sector comparison.Failed drives have more RS across all disk models.Many disks fail before theyexhaust their spare sectors.Failed drives with bigger capacity have more RS.Y-axis scales vary.the working and failed disk sets.The x-axis represents the Kth deciles,with the error counts on the y-axis.Reallocated sector :Figure 6presents the number of RS on failed and working drives.We observe that the ma-jority of failed drives developed a large number of RS while most that are working have only a few.For ex-ample,80%of A-2failed drives have more than 23RS but 90%of working drives have less than 29of this er-ror.Every disk model demonstrates a similar pattern;the only difference is how large the discrimination is.Failed disks have different RS counts,implying that many disks fail before they use up all spare sectors.We also find that failed drives with bigger capacity tend to have more RS,though the numbers depend more on the maximum number of reallocations permitted than the total size.For example,the median count of RS on A-2failed drives is 327,compared to 171for A-1;A-2has both twice the ca-pacity and twice the maximum number of reallocations,so this difference is expected.On the other hand,C-2has twice the capacity as C-1but the same maximum num-ber of RS (2048),and its 9th decile of RS is only 40%higher than C-1.(Note that the median RS count for C-1is zero,implying that many C-1disks fail for reasons other than reallocated sectors;this is consistent with the large infant mortality shown in Figure 4and bears further investigation.D-1has similar characteristics.)Media error :Due to the limitation of the logging mes-sages we have on hand,we can analyze this error type only on the A-2disk model.The result is presented in Figure 7.Though failed disks have more media errors than working ones,the discrimination is not that signif-0102030405060708090100110M e d i u m E r r o r C o u n t123456789Deciles2359152232478611123471330failed disk deciles working disk decilesA-2Figure 7:Media error comparison.There is only mod-erate discrimination.Shown only for A-2.icant compared to RS.For example,50%of failed disks have fewer than 15media errors,and 50%of working ones developed more than 3errors.There is a large over-lap between them,perhaps because only sector errors de-tected in read operations are reported as media errors.Sector errors detected in writes will trigger the reallo-cation process directly without notifying the upper layer.Since the RAID layer will re-write the reconstructed data upon a detected media error,which causes the realloca-tion process,every media error will lead to an RS even-tually:the media error count is thus a subset of RS.More details can be found in §2.2.Pending and Uncorrectable sectors :As introduced in §2.1,sector errors discovered through the disk internal scan will be marked as pending sectors or uncorrectable sectors.The results for pending sectors are presented。
Missing Value Imputation Based on Data Clustering
Missing values may appear either in conditional attributes or in class attribute (target attribute). There are many approaches to deal with missing values described in [6], for instance: (a) Ignore objects containing missing values; (b) Fill the missing value manually; (c) Substitute the missing values by a global constant or the mean of the objects ; (d) Get the most probable value to fill in the missing values. The first approach usually lost too much useful information, whereas the second one is timeconsuming and expensive in cost, so it is infeasible in many applications. The third approach assumes that all missing values are with the same value, probably leading to considerable distortions in data distribution. However, Han et al. 2000, Zhang et al. 2005 in [2, 6] think: ‘The method of imputation, however, is a popular strategy. In comparison to other methods, it uses as m ore information as possible from the observed data to predict missing values . Traditional missing value imputation techniques can be roughly classified into parametric imputation (e.g., the linear regression) and non-parametric imputation (e.g., non-parametric kernel-based regression method [20, 21, 22], Nearest Neighbor method [4, 6] (referred to as NN)). The parametric regression imputation is superior if a dataset can be adequately modeled parametrically, or if users can correctly specify the parametric forms for the dataset. For instance, the linear regression methods usually can treat well the continuous target attribute, which is a linear combination of the conditional attributes. However, when we don’t know the actual relation between the conditional attributes and the target attribute, the performance of the linear regression for imp uting missing values is very poor. In real application, if the model is misspecified (in fact, it is usually impossible for us to know the distribution of the real dataset), the estimations of parametric method may be highly biased and the optimal control factor settings may be miscalculated. Non-parametric imputation algorithm, which can provide superior fit by capturing structure in the dataset (note that a misspecified parametric model cannot), offers a nice alternative if users have no idea on the actual distribution of a dataset. For example, the NN method is regarded as one of non-parametric techniques used to compensate for missing values in sample surveys [7]. And it has been successfully used in, for instance, U.S. Census Bureau and Canadian Census Bureau. What’s more, using a non-parametric algorithm is beneficial when the form of relationship between the conditional attributes and the target attribute is not known a-priori [8]. While nonparametric imputation method is of low-efficiency, the popular NN method faces two issues: (1) each instance with missing values requires the calculation of the distances from it to all other instances in a dataset; and (2) there are only a few random chances for selecting the nearest neighbor. This paper addresses the above issues by proposing a clustering-based non-parametric regression method for dealing with the problem of missing value in target attribute (named Clusteringbased Missing value Imputation, denoted as CMI). In our approach, we fill up the missing values with plausible values that are generated by using a kernel-based method. Specifically, we first divide the dataset (including instances with missing values) into clusters. Then each instance with missing-values is assigned to a cluster most similar to it. Finally, missing values of an instance A are patched up with the plausible values generated from A’s cluster. The rest of the paper is organized as follows. In section 2, we give related work on missing values imputation. Section 3 presents our method in detail. Extensive
稀疏恢复和傅里叶采样
Accepted by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Leslie A. Kolodziejski Chair, Department Committee on Graduate Students
2
Sparse Recovery and Fourier Sampling by Eric Price
Submitted to the Department of Electrical Engineering and Computer Science on August 26, 2013, in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Computer Science
Author . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Department of Electrical Engineering and Computer Science August 26, 2013
Certified by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Piotr Indyk Professor Thesis Supervisor
ACM-GIS%202006-A%20Peer-to-Peer%20Spatial%20Cloaking%20Algorithm%20for%20Anonymous%20Location-based%
A Peer-to-Peer Spatial Cloaking Algorithm for AnonymousLocation-based Services∗Chi-Yin Chow Department of Computer Science and Engineering University of Minnesota Minneapolis,MN cchow@ Mohamed F.MokbelDepartment of ComputerScience and EngineeringUniversity of MinnesotaMinneapolis,MNmokbel@Xuan LiuIBM Thomas J.WatsonResearch CenterHawthorne,NYxuanliu@ABSTRACTThis paper tackles a major privacy threat in current location-based services where users have to report their ex-act locations to the database server in order to obtain their desired services.For example,a mobile user asking about her nearest restaurant has to report her exact location.With untrusted service providers,reporting private location in-formation may lead to several privacy threats.In this pa-per,we present a peer-to-peer(P2P)spatial cloaking algo-rithm in which mobile and stationary users can entertain location-based services without revealing their exact loca-tion information.The main idea is that before requesting any location-based service,the mobile user will form a group from her peers via single-hop communication and/or multi-hop routing.Then,the spatial cloaked area is computed as the region that covers the entire group of peers.Two modes of operations are supported within the proposed P2P spa-tial cloaking algorithm,namely,the on-demand mode and the proactive mode.Experimental results show that the P2P spatial cloaking algorithm operated in the on-demand mode has lower communication cost and better quality of services than the proactive mode,but the on-demand incurs longer response time.Categories and Subject Descriptors:H.2.8[Database Applications]:Spatial databases and GISGeneral Terms:Algorithms and Experimentation. Keywords:Mobile computing,location-based services,lo-cation privacy and spatial cloaking.1.INTRODUCTIONThe emergence of state-of-the-art location-detection de-vices,e.g.,cellular phones,global positioning system(GPS) devices,and radio-frequency identification(RFID)chips re-sults in a location-dependent information access paradigm,∗This work is supported in part by the Grants-in-Aid of Re-search,Artistry,and Scholarship,University of Minnesota. Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on thefirst page.To copy otherwise,to republish,to post on servers or to redistribute to lists,requires prior specific permission and/or a fee.ACM-GIS’06,November10-11,2006,Arlington,Virginia,USA. Copyright2006ACM1-59593-529-0/06/0011...$5.00.known as location-based services(LBS)[30].In LBS,mobile users have the ability to issue location-based queries to the location-based database server.Examples of such queries include“where is my nearest gas station”,“what are the restaurants within one mile of my location”,and“what is the traffic condition within ten minutes of my route”.To get the precise answer of these queries,the user has to pro-vide her exact location information to the database server. With untrustworthy servers,adversaries may access sensi-tive information about specific individuals based on their location information and issued queries.For example,an adversary may check a user’s habit and interest by knowing the places she visits and the time of each visit,or someone can track the locations of his ex-friends.In fact,in many cases,GPS devices have been used in stalking personal lo-cations[12,39].To tackle this major privacy concern,three centralized privacy-preserving frameworks are proposed for LBS[13,14,31],in which a trusted third party is used as a middleware to blur user locations into spatial regions to achieve k-anonymity,i.e.,a user is indistinguishable among other k−1users.The centralized privacy-preserving frame-work possesses the following shortcomings:1)The central-ized trusted third party could be the system bottleneck or single point of failure.2)Since the centralized third party has the complete knowledge of the location information and queries of all users,it may pose a serious privacy threat when the third party is attacked by adversaries.In this paper,we propose a peer-to-peer(P2P)spatial cloaking algorithm.Mobile users adopting the P2P spatial cloaking algorithm can protect their privacy without seeking help from any centralized third party.Other than the short-comings of the centralized approach,our work is also moti-vated by the following facts:1)The computation power and storage capacity of most mobile devices have been improv-ing at a fast pace.2)P2P communication technologies,such as IEEE802.11and Bluetooth,have been widely deployed.3)Many new applications based on P2P information shar-ing have rapidly taken shape,e.g.,cooperative information access[9,32]and P2P spatio-temporal query processing[20, 24].Figure1gives an illustrative example of P2P spatial cloak-ing.The mobile user A wants tofind her nearest gas station while beingfive anonymous,i.e.,the user is indistinguish-able amongfive users.Thus,the mobile user A has to look around andfind other four peers to collaborate as a group. In this example,the four peers are B,C,D,and E.Then, the mobile user A cloaks her exact location into a spatialA B CDEBase Stationregion that covers the entire group of mobile users A ,B ,C ,D ,and E .The mobile user A randomly selects one of the mobile users within the group as an agent .In the ex-ample given in Figure 1,the mobile user D is selected as an agent.Then,the mobile user A sends her query (i.e.,what is the nearest gas station)along with her cloaked spa-tial region to the agent.The agent forwards the query to the location-based database server through a base station.Since the location-based database server processes the query based on the cloaked spatial region,it can only give a list of candidate answers that includes the actual answers and some false positives.After the agent receives the candidate answers,it forwards the candidate answers to the mobile user A .Finally,the mobile user A gets the actual answer by filtering out all the false positives.The proposed P2P spatial cloaking algorithm can operate in two modes:on-demand and proactive .In the on-demand mode,mobile clients execute the cloaking algorithm when they need to access information from the location-based database server.On the other side,in the proactive mode,mobile clients periodically look around to find the desired number of peers.Thus,they can cloak their exact locations into spatial regions whenever they want to retrieve informa-tion from the location-based database server.In general,the contributions of this paper can be summarized as follows:1.We introduce a distributed system architecture for pro-viding anonymous location-based services (LBS)for mobile users.2.We propose the first P2P spatial cloaking algorithm for mobile users to entertain high quality location-based services without compromising their privacy.3.We provide experimental evidence that our proposed algorithm is efficient in terms of the response time,is scalable to large numbers of mobile clients,and is effective as it provides high-quality services for mobile clients without the need of exact location information.The rest of this paper is organized as follows.Section 2highlights the related work.The system model of the P2P spatial cloaking algorithm is presented in Section 3.The P2P spatial cloaking algorithm is described in Section 4.Section 5discusses the integration of the P2P spatial cloak-ing algorithm with privacy-aware location-based database servers.Section 6depicts the experimental evaluation of the P2P spatial cloaking algorithm.Finally,Section 7con-cludes this paper.2.RELATED WORKThe k -anonymity model [37,38]has been widely used in maintaining privacy in databases [5,26,27,28].The main idea is to have each tuple in the table as k -anonymous,i.e.,indistinguishable among other k −1tuples.Although we aim for the similar k -anonymity model for the P2P spatial cloaking algorithm,none of these techniques can be applied to protect user privacy for LBS,mainly for the following four reasons:1)These techniques preserve the privacy of the stored data.In our model,we aim not to store the data at all.Instead,we store perturbed versions of the data.Thus,data privacy is managed before storing the data.2)These approaches protect the data not the queries.In anonymous LBS,we aim to protect the user who issues the query to the location-based database server.For example,a mobile user who wants to ask about her nearest gas station needs to pro-tect her location while the location information of the gas station is not protected.3)These approaches guarantee the k -anonymity for a snapshot of the database.In LBS,the user location is continuously changing.Such dynamic be-havior calls for continuous maintenance of the k -anonymity model.(4)These approaches assume a unified k -anonymity requirement for all the stored records.In our P2P spatial cloaking algorithm,k -anonymity is a user-specified privacy requirement which may have a different value for each user.Motivated by the privacy threats of location-detection de-vices [1,4,6,40],several research efforts are dedicated to protect the locations of mobile users (e.g.,false dummies [23],landmark objects [18],and location perturbation [10,13,14]).The most closed approaches to ours are two centralized spatial cloaking algorithms,namely,the spatio-temporal cloaking [14]and the CliqueCloak algorithm [13],and one decentralized privacy-preserving algorithm [23].The spatio-temporal cloaking algorithm [14]assumes that all users have the same k -anonymity requirements.Furthermore,it lacks the scalability because it deals with each single request of each user individually.The CliqueCloak algorithm [13]as-sumes a different k -anonymity requirement for each user.However,since it has large computation overhead,it is lim-ited to a small k -anonymity requirement,i.e.,k is from 5to 10.A decentralized privacy-preserving algorithm is proposed for LBS [23].The main idea is that the mobile client sends a set of false locations,called dummies ,along with its true location to the location-based database server.However,the disadvantages of using dummies are threefold.First,the user has to generate realistic dummies to pre-vent the adversary from guessing its true location.Second,the location-based database server wastes a lot of resources to process the dummies.Finally,the adversary may esti-mate the user location by using cellular positioning tech-niques [34],e.g.,the time-of-arrival (TOA),the time differ-ence of arrival (TDOA)and the direction of arrival (DOA).Although several existing distributed group formation al-gorithms can be used to find peers in a mobile environment,they are not designed for privacy preserving in LBS.Some algorithms are limited to only finding the neighboring peers,e.g.,lowest-ID [11],largest-connectivity (degree)[33]and mobility-based clustering algorithms [2,25].When a mo-bile user with a strict privacy requirement,i.e.,the value of k −1is larger than the number of neighboring peers,it has to enlist other peers for help via multi-hop routing.Other algorithms do not have this limitation,but they are designed for grouping stable mobile clients together to facil-Location-based Database ServerDatabase ServerDatabase ServerFigure 2:The system architectureitate efficient data replica allocation,e.g.,dynamic connec-tivity based group algorithm [16]and mobility-based clus-tering algorithm,called DRAM [19].Our work is different from these approaches in that we propose a P2P spatial cloaking algorithm that is dedicated for mobile users to dis-cover other k −1peers via single-hop communication and/or via multi-hop routing,in order to preserve user privacy in LBS.3.SYSTEM MODELFigure 2depicts the system architecture for the pro-posed P2P spatial cloaking algorithm which contains two main components:mobile clients and location-based data-base server .Each mobile client has its own privacy profile that specifies its desired level of privacy.A privacy profile includes two parameters,k and A min ,k indicates that the user wants to be k -anonymous,i.e.,indistinguishable among k users,while A min specifies the minimum resolution of the cloaked spatial region.The larger the value of k and A min ,the more strict privacy requirements a user needs.Mobile users have the ability to change their privacy profile at any time.Our employed privacy profile matches the privacy re-quirements of mobiles users as depicted by several social science studies (e.g.,see [4,15,17,22,29]).In this architecture,each mobile user is equipped with two wireless network interface cards;one of them is dedicated to communicate with the location-based database server through the base station,while the other one is devoted to the communication with other peers.A similar multi-interface technique has been used to implement IP multi-homing for stream control transmission protocol (SCTP),in which a machine is installed with multiple network in-terface cards,and each assigned a different IP address [36].Similarly,in mobile P2P cooperation environment,mobile users have a network connection to access information from the server,e.g.,through a wireless modem or a base station,and the mobile users also have the ability to communicate with other peers via a wireless LAN,e.g.,IEEE 802.11or Bluetooth [9,24,32].Furthermore,each mobile client is equipped with a positioning device, e.g.,GPS or sensor-based local positioning systems,to determine its current lo-cation information.4.P2P SPATIAL CLOAKINGIn this section,we present the data structure and the P2P spatial cloaking algorithm.Then,we describe two operation modes of the algorithm:on-demand and proactive .4.1Data StructureThe entire system area is divided into grid.The mobile client communicates with each other to discover other k −1peers,in order to achieve the k -anonymity requirement.TheAlgorithm 1P2P Spatial Cloaking:Request Originator m 1:Function P2PCloaking-Originator (h ,k )2://Phase 1:Peer searching phase 3:The hop distance h is set to h4:The set of discovered peers T is set to {∅},and the number ofdiscovered peers k =|T |=05:while k <k −1do6:Broadcast a FORM GROUP request with the parameter h (Al-gorithm 2gives the response of each peer p that receives this request)7:T is the set of peers that respond back to m by executingAlgorithm 28:k =|T |;9:if k <k −1then 10:if T =T then 11:Suspend the request 12:end if 13:h ←h +1;14:T ←T ;15:end if 16:end while17://Phase 2:Location adjustment phase 18:for all T i ∈T do19:|mT i .p |←the greatest possible distance between m and T i .pby considering the timestamp of T i .p ’s reply and maximum speed20:end for21://Phase 3:Spatial cloaking phase22:Form a group with k −1peers having the smallest |mp |23:h ←the largest hop distance h p of the selected k −1peers 24:Determine a grid area A that covers the entire group 25:if A <A min then26:Extend the area of A till it covers A min 27:end if28:Randomly select a mobile client of the group as an agent 29:Forward the query and A to the agentmobile client can thus blur its exact location into a cloaked spatial region that is the minimum grid area covering the k −1peers and itself,and satisfies A min as well.The grid area is represented by the ID of the left-bottom and right-top cells,i.e.,(l,b )and (r,t ).In addition,each mobile client maintains a parameter h that is the required hop distance of the last peer searching.The initial value of h is equal to one.4.2AlgorithmFigure 3gives a running example for the P2P spatial cloaking algorithm.There are 15mobile clients,m 1to m 15,represented as solid circles.m 8is the request originator,other black circles represent the mobile clients received the request from m 8.The dotted circles represent the commu-nication range of the mobile client,and the arrow represents the movement direction.Algorithms 1and 2give the pseudo code for the request originator (denoted as m )and the re-quest receivers (denoted as p ),respectively.In general,the algorithm consists of the following three phases:Phase 1:Peer searching phase .The request origina-tor m wants to retrieve information from the location-based database server.m first sets h to h ,a set of discovered peers T to {∅}and the number of discovered peers k to zero,i.e.,|T |.(Lines 3to 4in Algorithm 1).Then,m broadcasts a FORM GROUP request along with a message sequence ID and the hop distance h to its neighboring peers (Line 6in Algorithm 1).m listens to the network and waits for the reply from its neighboring peers.Algorithm 2describes how a peer p responds to the FORM GROUP request along with a hop distance h and aFigure3:P2P spatial cloaking algorithm.Algorithm2P2P Spatial Cloaking:Request Receiver p1:Function P2PCloaking-Receiver(h)2://Let r be the request forwarder3:if the request is duplicate then4:Reply r with an ACK message5:return;6:end if7:h p←1;8:if h=1then9:Send the tuple T=<p,(x p,y p),v maxp ,t p,h p>to r10:else11:h←h−1;12:Broadcast a FORM GROUP request with the parameter h 13:T p is the set of peers that respond back to p14:for all T i∈T p do15:T i.h p←T i.h p+1;16:end for17:T p←T p∪{<p,(x p,y p),v maxp ,t p,h p>};18:Send T p back to r19:end ifmessage sequence ID from another peer(denoted as r)that is either the request originator or the forwarder of the re-quest.First,p checks if it is a duplicate request based on the message sequence ID.If it is a duplicate request,it sim-ply replies r with an ACK message without processing the request.Otherwise,p processes the request based on the value of h:Case1:h= 1.p turns in a tuple that contains its ID,current location,maximum movement speed,a timestamp and a hop distance(it is set to one),i.e.,< p,(x p,y p),v max p,t p,h p>,to r(Line9in Algorithm2). Case2:h> 1.p decrements h and broadcasts the FORM GROUP request with the updated h and the origi-nal message sequence ID to its neighboring peers.p keeps listening to the network,until it collects the replies from all its neighboring peers.After that,p increments the h p of each collected tuple,and then it appends its own tuple to the collected tuples T p.Finally,it sends T p back to r (Lines11to18in Algorithm2).After m collects the tuples T from its neighboring peers, if m cannotfind other k−1peers with a hop distance of h,it increments h and re-broadcasts the FORM GROUP request along with a new message sequence ID and h.m repeatedly increments h till itfinds other k−1peers(Lines6to14in Algorithm1).However,if mfinds the same set of peers in two consecutive broadcasts,i.e.,with hop distances h and h+1,there are not enough connected peers for m.Thus, m has to relax its privacy profile,i.e.,use a smaller value of k,or to be suspended for a period of time(Line11in Algorithm1).Figures3(a)and3(b)depict single-hop and multi-hop peer searching in our running example,respectively.In Fig-ure3(a),the request originator,m8,(e.g.,k=5)canfind k−1peers via single-hop communication,so m8sets h=1. Since h=1,its neighboring peers,m5,m6,m7,m9,m10, and m11,will not further broadcast the FORM GROUP re-quest.On the other hand,in Figure3(b),m8does not connect to k−1peers directly,so it has to set h>1.Thus, its neighboring peers,m7,m10,and m11,will broadcast the FORM GROUP request along with a decremented hop dis-tance,i.e.,h=h−1,and the original message sequence ID to their neighboring peers.Phase2:Location adjustment phase.Since the peer keeps moving,we have to capture the movement between the time when the peer sends its tuple and the current time. For each received tuple from a peer p,the request originator, m,determines the greatest possible distance between them by an equation,|mp |=|mp|+(t c−t p)×v max p,where |mp|is the Euclidean distance between m and p at time t p,i.e.,|mp|=(x m−x p)2+(y m−y p)2,t c is the currenttime,t p is the timestamp of the tuple and v maxpis the maximum speed of p(Lines18to20in Algorithm1).In this paper,a conservative approach is used to determine the distance,because we assume that the peer will move with the maximum speed in any direction.If p gives its movement direction,m has the ability to determine a more precise distance between them.Figure3(c)illustrates that,for each discovered peer,the circle represents the largest region where the peer can lo-cate at time t c.The greatest possible distance between the request originator m8and its discovered peer,m5,m6,m7, m9,m10,or m11is represented by a dotted line.For exam-ple,the distance of the line m8m 11is the greatest possible distance between m8and m11at time t c,i.e.,|m8m 11|. Phase3:Spatial cloaking phase.In this phase,the request originator,m,forms a virtual group with the k−1 nearest peers,based on the greatest possible distance be-tween them(Line22in Algorithm1).To adapt to the dynamic network topology and k-anonymity requirement, m sets h to the largest value of h p of the selected k−1 peers(Line15in Algorithm1).Then,m determines the minimum grid area A covering the entire group(Line24in Algorithm1).If the area of A is less than A min,m extends A,until it satisfies A min(Lines25to27in Algorithm1). Figure3(c)gives the k−1nearest peers,m6,m7,m10,and m11to the request originator,m8.For example,the privacy profile of m8is(k=5,A min=20cells),and the required cloaked spatial region of m8is represented by a bold rectan-gle,as depicted in Figure3(d).To issue the query to the location-based database server anonymously,m randomly selects a mobile client in the group as an agent(Line28in Algorithm1).Then,m sendsthe query along with the cloaked spatial region,i.e.,A,to the agent(Line29in Algorithm1).The agent forwards thequery to the location-based database server.After the serverprocesses the query with respect to the cloaked spatial re-gion,it sends a list of candidate answers back to the agent.The agent forwards the candidate answer to m,and then mfilters out the false positives from the candidate answers. 4.3Modes of OperationsThe P2P spatial cloaking algorithm can operate in twomodes,on-demand and proactive.The on-demand mode:The mobile client only executesthe algorithm when it needs to retrieve information from the location-based database server.The algorithm operatedin the on-demand mode generally incurs less communica-tion overhead than the proactive mode,because the mobileclient only executes the algorithm when necessary.However,it suffers from a longer response time than the algorithm op-erated in the proactive mode.The proactive mode:The mobile client adopting theproactive mode periodically executes the algorithm in back-ground.The mobile client can cloak its location into a spa-tial region immediately,once it wants to communicate withthe location-based database server.The proactive mode pro-vides a better response time than the on-demand mode,but it generally incurs higher communication overhead and giveslower quality of service than the on-demand mode.5.ANONYMOUS LOCATION-BASEDSERVICESHaving the spatial cloaked region as an output form Algo-rithm1,the mobile user m sends her request to the location-based server through an agent p that is randomly selected.Existing location-based database servers can support onlyexact point locations rather than cloaked regions.In or-der to be able to work with a spatial region,location-basedservers need to be equipped with a privacy-aware queryprocessor(e.g.,see[29,31]).The main idea of the privacy-aware query processor is to return a list of candidate answerrather than the exact query answer.Then,the mobile user m willfilter the candidate list to eliminate its false positives andfind its exact answer.The tighter the spatial cloaked re-gion,the lower is the size of the candidate answer,and hencethe better is the performance of the privacy-aware query processor.However,tight cloaked regions may represent re-laxed privacy constrained.Thus,a trade-offbetween the user privacy and the quality of service can be achieved[31]. Figure4(a)depicts such scenario by showing the data stored at the server side.There are32target objects,i.e., gas stations,T1to T32represented as black circles,the shaded area represents the spatial cloaked area of the mo-bile client who issued the query.For clarification,the actual mobile client location is plotted in Figure4(a)as a black square inside the cloaked area.However,such information is neither stored at the server side nor revealed to the server. The privacy-aware query processor determines a range that includes all target objects that are possibly contributing to the answer given that the actual location of the mobile client could be anywhere within the shaded area.The range is rep-resented as a bold rectangle,as depicted in Figure4(b).The server sends a list of candidate answers,i.e.,T8,T12,T13, T16,T17,T21,and T22,back to the agent.The agent next for-(a)Server Side(b)Client SideFigure4:Anonymous location-based services wards the candidate answers to the requesting mobile client either through single-hop communication or through multi-hop routing.Finally,the mobile client can get the actualanswer,i.e.,T13,byfiltering out the false positives from thecandidate answers.The algorithmic details of the privacy-aware query proces-sor is beyond the scope of this paper.Interested readers are referred to[31]for more details.6.EXPERIMENTAL RESULTSIn this section,we evaluate and compare the scalabilityand efficiency of the P2P spatial cloaking algorithm in boththe on-demand and proactive modes with respect to the av-erage response time per query,the average number of mes-sages per query,and the size of the returned candidate an-swers from the location-based database server.The queryresponse time in the on-demand mode is defined as the timeelapsed between a mobile client starting to search k−1peersand receiving the candidate answers from the agent.On theother hand,the query response time in the proactive mode is defined as the time elapsed between a mobile client startingto forward its query along with the cloaked spatial regionto the agent and receiving the candidate answers from theagent.The simulation model is implemented in C++usingCSIM[35].In all the experiments in this section,we consider an in-dividual random walk model that is based on“random way-point”model[7,8].At the beginning,the mobile clientsare randomly distributed in a spatial space of1,000×1,000square meters,in which a uniform grid structure of100×100cells is constructed.Each mobile client randomly chooses itsown destination in the space with a randomly determined speed s from a uniform distribution U(v min,v max).When the mobile client reaches the destination,it comes to a stand-still for one second to determine its next destination.Afterthat,the mobile client moves towards its new destinationwith another speed.All the mobile clients repeat this move-ment behavior during the simulation.The time interval be-tween two consecutive queries generated by a mobile client follows an exponential distribution with a mean of ten sec-onds.All the experiments consider one half-duplex wirelesschannel for a mobile client to communicate with its peers with a total bandwidth of2Mbps and a transmission range of250meters.When a mobile client wants to communicate with other peers or the location-based database server,it has to wait if the requested channel is busy.In the simulated mobile environment,there is a centralized location-based database server,and one wireless communication channel between the location-based database server and the mobile。
基于线性规划的穿越沙漠游戏通关研究 赵凯龙
基于线性规划的穿越沙漠游戏通关研究赵凯龙发表时间：2020-11-03T14:33:12.040Z 来源：《论证与研究》2020年9期作者：赵凯龙  杜 庚Q [导读] 摘要：本文主要针对解决穿越沙漠游戏的相关问题的研究。
利用图论中Floyd算法，借助MATLAB软件求解得所需要的最短路径，做了多种情形下的不同地图模式，对穿越沙漠的行程安排问题进行了探究。
首先基于沙漠实际情况建立相关的线性规划模型求解收益最优化问题。
其次使用0-1整数规划对各天气下的具体情况进行独立分析，确定最短路时，将各区域画点建立邻接矩阵赋权图。
最后对未知的天气情况进行预测，得到三种天气出现的赵凯龙 杜 庚 臧玉婷 王乐宁 李思琦（青岛理工大学 山东 青岛 266500）摘要：本文主要针对解决穿越沙漠游戏的相关问题的研究。
利用图论中Floyd算法，借助MATLAB软件求解得所需要的最短路径，做了多种情形下的不同地图模式，对穿越沙漠的行程安排问题进行了探究。
首先基于沙漠实际情况建立相关的线性规划模型求解收益最优化问题。
其次使用0-1整数规划对各天气下的具体情况进行独立分析，确定最短路时，将各区域画点建立邻接矩阵赋权图。
最后对未知的天气情况进行预测，得到三种天气出现的概率分布，将其作为权值对消耗资源对应的资金数进行加权，将各路线上的天气资源消耗量进行加权，得到线路完成后的剩余资金量，对其得到的值进行比较以可以确定最优方案。
关键词：最短路径；线性规划最优化；邻接矩阵引言虚拟的一切皆源于现实，游戏中物理空间采用现实地理背景，扩展了人所在的空间，从而为人类的发展提供实验借鉴价值，并对现实物理世界起启示作用[1]。
基于以上背景完成游戏，最终目标是在规定时间内到达终点并保留最大资金数。
游戏成功或失败。
游戏从第0天开始，每天消耗不同数量的水和食物。
若玩家在截止日期或之前到达终点即为成功，若资源耗尽或未达到终点则为失败。
Visibility culling using hierarchical occlusion maps
Visibility Culling using Hierarchical Occlusion Maps Hansong Zhang Dinesh Manocha Tom Hudson Kenneth E.Hoff IIIDepartment of Computer ScienceUniversity of North CarolinaChapel Hill,NC27599-3175zhangh,dm,hudson,hoff@/˜zhangh,dm,hudson,hoffAbstract:We present hierarchical occlusion maps(HOM)for visibility culling on complex models with high depth complexity. The culling algorithm uses an object space bounding volume hier-archy and a hierarchy of image space occlusion maps.Occlusion maps represent the aggregate of projections of the occluders onto the image plane.For each frame,the algorithm selects a small set of objects from the model as occludersand renders them to form an initial occlusion map,from which a hierarchy of occlusion maps is built.The occlusion maps are used to cull away a portion of the model not visible from the current viewpoint.The algorithm is applicable to all models and makes no assumptions about the size,shape,or type of occluders.It supports approximate culling in which small holes in or among occluders can be ignored.The algorithm has been implemented on current graphics systems and has been applied to large models composed of hundreds of thou-sands of polygons.In practice,it achieves significant speedup in interactive walkthroughs of models with high depth complexity. CR Categories and Subject Descriptors:I.3.5[Computer Graphics]:Computational Geometry and Object Modeling Key Words and Phrases:visibility culling,interactive display, image pyramid,occlusion culling,hierarchical data structures1IntroductionInteractive display and walkthrough of large geometric models currently pushes the limits of graphics technology.Environments composed of millions of primitives(e.g.polygons)are not un-common in applications such as simulation-based design of large mechanical systems,architectural visualization,or walkthrough of outdoor scenes.Although throughput of graphics systems has increased considerably over the years,the size and complexity of these environments have been growing even faster.In order to display such models at interactive rates,the rendering algorithms need to use techniques based on visibility culling,levels-of-detail, texturing,etc.to limit the number of primitives rendered in each frame.In this paper,we focus on visibility culling algorithms, whose goal is to cull away large portions of the environment not visible from the current viewpoint.Our criteria for an effective visibility culling algorithm are gen-erality,interactive performance,and significant culling.Addi-tionally,in order for it to be practical,it should be implementable on current graphics systems and work well on large real-world models.Main Contribution:In this paper,we present a new algorithm for visibility culling in complex environments with highdepth Figure1:Demonstration of our algorithm on the CAD model of a submarine’s auxiliary machine room.The model has632,252 polygons.The green lines outline the viewing frustum.Blue indicates objects selected as occluders,gray the objects not culled by our algorithm and transparent red the objects culled away.For this particular view,82.7%of the model is culled. complexity.At each frame,the algorithm carefully selects a small subset of the model as occluders and renders the occluders to build hierarchical occlusion maps(HOM).The hierarchy is an image pyramid and each map in the hierarchy is composed of pixels corresponding to rectangular blocks in the screen space. The pixel value records the opacity of the block.The algorithm decomposesthe visibility test for an object into a two-dimensional overlap test,performed against the occlusion map hierarchy,and a conservative test to compare the depth.The overall approach combines an object space bounding volume hierarchy(also useful for view frustum culling)with the image space occlusion map hierarchy to cull away a portion of the model not visible from the current viewpoint.Some of the main features of the algorithm are:1.Generality:The algorithm requires no special structures inthe model and places no restriction on the types of occluders.The occluders may be polygonal objects,curved surfaces,or even not be geometrically defined(e.g.a billboard).2.Occluder Fusion:A key characteristic of the algorithm isthe ability to combine a“forest”of small or disjoint occlud-ers,rather than using only large occluders.In most cases,the union of a set of occluders can occlude much more than what each of them can occlude taken separately.This is very useful for large mechanical CAD and outdoor models.3.Significant Culling:On high depth complexity models,thealgorithm is able to cull away a significant fraction(up to 95%)of the model from most viewpoints.4.Portability:The algorithm can be implemented on mostcurrent graphics systems.Its main requirement is the ability to read back the frame-buffer.The construction of hierarchi-cal occlusion maps can be accelerated by texture mapping hardware.It is not susceptible to degeneracies in the input and can be parallelized on multiprocessor systems.5.Efficiency:The construction of occlusion maps takes afew milliseconds per frame on current medium-to high-end graphics systems.The culling algorithm achieves sig-nificant speedup in interactive walkthroughs of models with high depth complexity.The algorithm involves no significant preprocessing and is applicable to dynamic environments.6.Approximate Visibility Culling:Our approach can alsouse the hierarchy of maps to perform approximate culling.By varying an opacity threshold parameter the algorithm is able tofill small transparent holes in the occlusion maps and to cull away portions of the model which are visible through small gaps in the occluders.The resulting algorithm has been implemented on different platforms(SGI Max Impact and Infinite Reality)and applied to city models,CAD models,and dynamic environments.It ob-tains considerable speedup in overall frame rate.In Figure1we demonstrate its performance on a submarine’s Auxiliary Machine Room.Organization:The rest of the paper is organized as follows:.We briefly survey related work in Section2and give an overview of our approachin Section3.Section4describes occlusion maps and techniques for fast implementation on current graphics systems.In Section5we describe the entire culling algorithm.We describe its implementation and performance in Section6.Section7analyses our algorithm and compares it with other approaches.Finally,in Section8,we briefly describe some future directions.2Related WorkVisibility computation and hidden surface removal are classic problems in computer graphics[FDHF90].Some of the com-monly used visibility algorithms are based on-buffer[Cat74] and view-frustum culling[Cla76,GBW90].Others include Painter’s Algorithm[FDHF90]and area-subdivision algorithms [War69,FDHF90].There is significant literature on visible surface computa-tion in computational geometry.Many asymptotically effi-cient algorithms have been proposed for hidden surface removal [Mul89,McK87](see[Dor94]for a recent survey).However,the practical utility of these algorithms is unclear at the moment.Efficient algorithms for calculating the visibility relationship among a static group of3D polygons from arbitrary viewpoints have been proposed based on the binary space-partitioning(BSP) tree[FKN80].The tree construction may involve considerable pre-processing in terms of time and space requirements for large models.In[Nay92],Naylor has given an output-sensitive visi-bility algorithm using BSPs.It uses a2D BSP tree to represent images and presents an algorithm to project a3D BSP tree,repre-senting the model in object space,into a2D BSP tree representing its image.Many algorithms structure the model database into cells or re-gions,and use a combination of off-line and on-line algorithms for cell-to-cell visibility and the conservative computation of the potentially visible set(PVS)of primitives[ARB90,TS91,LG95]. Such approaches have been successfully used to visualize archi-tectural models,where the division of a building into discrete rooms lends itself to a natural division of the database into cells. It is not apparent that cell-based approaches can be generalized to an arbitrary model.Other algorithms for densely-occluded but somewhat less-structured models have been proposed by Yagel and Ray[YR96]. They used regular spatial subdivision to partition the model into cells and describe a2D implementation.However,the resulting algorithm is very memory-intensive and does not scale well to large models.Object space algorithms for occlusion culling in general polygonal models have been presented by Coorg and Teller [CT96b,CT96a]and Hudson et al.[Hud96].These algorithms dynamically compute a subset of the objects as occluders and use them to cull away portions of the model.In particular, [CT96b,CT96a]compute an arrangement corresponding to a lin-earized portion of an aspect graph and track the viewpoint within it to check for occlusion.[Hud96]use shadow frusta and fast interference tests for occlusion culling.All of them are object-space algorithms and the choice of occluder is restricted to convex objects or simple combination of convex objects(e.g.two convex polytope sharing an edge).These algorithms are unable to com-bine a“forest"of small non-convex or disjoint occluders to cull away large portions of the model.A hierarchical Z-buffer algorithm combining spatial and tem-poral coherencehas been presented in[GKM93,GK94,Gre95].It uses two hierarchical data structures:an octree and a Z-pyramid. The algorithm exploits coherence by performing visibility queries on the Z-pyramid and is very effective in culling large portions of high-depth complexity models.However,most current graphics systems do not support the Z-pyramid capability in hardware,and simulating it in software can be relatively expensive.In[GK94], Greene and Kass used a quadtree data structure to test visibil-ity throughout image-space regions for anti-aliased rendering. [Geo95]describes an implementation of the Z-query operation on a parallel graphics architecture(PixelPlanes5)for obscuration culling.More recently,Greene[Gre96]has presented a hierarchical tiling algorithm using coverage masks.It uses an image hierarchy named a“coverage pyramid”for visibility culling.Traversing polygons from front to back,it can process densely occluded scenes efficiently and is well suited to anti-aliasing by oversam-pling andfiltering.For dynamic environments,Sudarsky and Gotsman[SG96] have presented an output-sensitive algorithm which minimizes the time required to update the hierarchical data structure for a dynamic object and minimize the number of dynamic objects for which the structure has to be updated.A number of techniques for interactive walkthrough of large geometric databases have been proposed.Refer to[RB96]for a recent survey.A number of commercial systems like Per-former[RH94],used for high performance graphics,and Brush [SBM94],used for visualizing architectural and CAD models, are available.They use techniques based on view-frustum culling, levels-of-detail,etc.,but have little support for occlusion culling on arbitrary models.There is substantial literature on the visibility problem from theflight simulator community.An overview offlight simulator architectures is given in[Mue95].Most notably,the Singer Com-pany’s Modular Digital Image Generator[Lat94]renders front to back using a hierarchy of mask buffers to skip over already cov-Figure2:Modified graphics pipeline showing our algorithm. The shaded blocks indicate components unique to culling with hierarchical occlusion map.ered spans,segments or rows in the image.General Electric’s COMPU-SCENE PT2000[Bun89]uses a similar algorithm but does not require the input polygons to be in front-to-back order and the mask buffer is not hierarchical.The Loral GT200[LORA]first renders near objects andfills in a mask buffer,which is used to cull away far objects.Sogitec’s APOGEE system uses the Meta-Z-buffer,which is similar to hierarchical Z buffer[Chu94].The structure of hierarchical occlusion maps is similar to some of the hierarchies that have been proposed for images,such as im-age pyramids[TP75],MIP maps[Wil83],-pyramids[GKM93], coverage pyramids[Gre96],and two-dimensional wavelet trans-forms like the non-standard decomposition[GBR91].3OverviewIn this paper we present a novel solution to the visibility problem. The heart of the algorithm is a hierarchy of occlusion maps,which records the aggregate projection of occluders onto the image plane at different resolutions.In other words,the maps capture the cu-mulative occluding effects of the occluders.We use occlusion maps because they can be built quickly and have several unique properties(described later in the paper).The use of occlusion maps reflects a decomposition of the visibility problem into two sub-problems:a two-dimensional overlap test and a depth test. The former decides whether the screen space projection of the po-tential occludee lies completely within the screen space projection of the union of all occluders.The latter determines whether or not the potential occludee is behind the occluders.We use occlusion maps for the overlap tests,and a depth estimation buffer for the conservative depth test.In the conventional-buffer algorithm (as well as in the hierarchical Z-buffer algorithm),the overlap test is implicitly performed as a side effect of the depth comparison by initializing the Z-buffer with large numbers.The algorithm renders the occluders at each frame and builds a hierarchy(pyramid)of occlusion maps.In addition to the model database,the algorithm maintains a separate occluder database, which is derived from the model database as a preprocessing step. Both databases are represented as bounding volume hierarchies. The rendering pipeline with our algorithm incorporated is illus-trated in Figure2.The shaded blocks indicate new stages intro-duced due to our algorithm.For each frame,the pipeline executes in two major phases:1.Construction of the Occlusion Map Hierarchy:The occluders are selected from the occluder database and rendered to build the occlusion map hierarchy.This involves:View-frustum culling:The algorithm traverses the bound-ing volume hierarchy of the occluder database tofind oc-cluders lying in the viewing frustum.Occluder selection:The algorithm selects a subset of the occluders lying in the viewing frustum.It utilizes temporal coherence between successive frames.Occluder rendering and depth estimation:The selected occluders are rendered to form an image in the framebuffer which is the highest resolution occlusion map.Objects are rendered in pure white with no lighting or texturing.The resulting image has only black and white pixels except for antialiased edges.A depth estimation buffer is built to record the depth of the occluders.Building the Hierarchical Occlusion Maps:After occlud-ers are rendered,the algorithm recursivelyfilters the ren-dered image down by averaging blocks of pixels.This pro-cess can be accelerated by texture mapping hardware on many current graphics systems.2.Visibility Culling with Hierarchical Occlusion Maps: Given an occlusion map hierarchy,the algorithm traverses the bounding volume hierarchy of the model database to perform visibility culling.The main components of this stage are: View-frustum Culling:The algorithm applies standard view-frustum culling to the model database.Depth Comparison:For each potential occludee,the algo-rithm conservatively checks whether it is behind the occlud-ers.Overlap test with Occlusion Maps:The algorithm tra-verses the occlusion map hierarchy to conservatively decide if each potential occludee’s screen space projection falls completely within the opaque areas of the maps.Only objects that fail one of the latter two tests(depth or over-lap)are rendered.4Occlusion MapsIn this section,we present occlusion maps,algorithms using tex-ture mapping hardware for fast construction of the hierarchy of occlusion maps,and state a number of properties of occlusion maps which are used by the visibility culling algorithm.When an opaque object is projected to the screen,the area of its projection is made opaque.The opacity of a block on the screen is defined as the ratio of the sum of the opaque areas in the block to the total area of the block.An occlusion map is a two-dimensional array in which each pixel records the opacity of a rectangularblock of screen space.Any rendered image can have an accompanying occlusion map which has the same resolution and stores the opacity for each pixel.In such a case,the occlusion map is essentially the channel[FDHF90]of the rendered image(assuming values for objects are set properly during rendering),though generally speaking a pixel in the occlusion map can correspond to a block of pixels in screen space.4.1Image PyramidGiven the lowest level occlusion map,the algorithm constructs from it a hierarchy of occlusion maps(HOM)by recursively ap-plying the average operator to rectangular blocks of pixels.This operation forms an image pyramid as shown in Figure3.The resulting hierarchy represents the occlusion map at multiple res-olutions.It greatly accelerates the overlap test and is used for approximate culling.In the rest of the paper,we follow the con-vention that the highest resolution occlusion map of a hierarchy is at level0.Figure3:The hierarchy of occlusion maps.This particular hierarchy is created by recursively averaging over2blocks of pixels.The outlined square marks the correspondence of one top-level pixel to pixels in the other levels.The image also shows the rendering of the torus to which the hierarchy corresponds.The algorithmfirst renders the occluders into an image,which forms the lowest-level and highest resolution occlusion map.This image represents an image-space fusion of all occluders in the object space.The occlusion map hierarchy is built by recursively filtering from the highest-resolution map down to some minimal resolution(e.g.44).The highest resolution need not match that of the image of the model ing a lower image resolution for rendering occluders may lead to inaccuracy for occlusion culling near the edges of the objects,but it speeds up the time for constructing the hierarchy.Furthermore,if hardware multi-sampled anti-aliasing is available,the lowest-level occlusion map has more accuracy.This is due to the fact that the anti-aliased image in itself is already afiltered down version of a larger super-sampled image on which the occluders were rendered.4.2Fast Construction of the Hierarchy Whenfiltering is performed on22blocks of pixels,hierarchy construction can be accelerated by graphics hardware that supports bilinear interpolation of texture maps.The averaging operator for 22blocks is actually a special case of bilinear interpolation. More precisely,the bilinear interpolation of four scalars or vectors 0123is:11011213where01,01are the weights.In our case,we use 05and this formula produces the average of the four values.By carefully setting the texture coordinates,we canfilter a 22occlusion map to by drawing a two dimensional rectangle of size,texturing it with the22occlusion map,and reading back the rendered image as the occlusion map.Figure4illustrates this process.The graphics hardware typically needs some setup time for the required operations.When the size of the map to befiltered is relatively small,setup time may dominate the computation.In such cases,the use of texture mapping hardware may slow down the computation of occlusion maps rather than accelerate it,and hierarchy building is faster on the host CPU.The break-even point between hardware and software hierarchy construction varies with different graphics systems.[BM96]presents a technique for generating mipmaps by using a hardware accumulation buffer.We did not use this method because the accumulation buffer is less commonly supported in current graphics systems than texture mapping.4.3Properties of Occlusion MapsThe hierarchical occlusion maps for an occluder set have sev-eral desirable properties for accelerating visibility culling.The visibility culling algorithm presented in Section5utilizes these properties.1.Occluder fusion:Occlusion maps represent the fusion of small and possibly disjoint occluders.No assumptions are made on the shape,size,or geometry of the occluders.Any object that is renderable can serve as an occluder.2.Hierarchical overlap test:The hierarchy allows us to perform a fast overlap test in screen space for visibility culling. This test is described in more detail in Section5.1.3.High-level opacity estimation:The opacity values in a low-resolution occlusion map give an estimate of the opacity values in higher-resolution maps.For instance,if a pixel in a higher level map has a very low intensity value,it implies that almost all of its descendant pixels have low opacities,i.e.there is a low possibility of occlusion.This is due to the fact that occlusion maps are based on the average operator rather than the minimum or maximum operators.This property allows for a conservative early termination of the overlap test.The opacity hierarchy also provides a natural method for ag-gressive early termination,or approximate occlusion culling.It may be used to cull away portions of the model visible only through small gaps in or among occluders.A high opacity value of a pixel in a low resolution map implies that most of its descendant pixels are opaque.The algorithm uses the opacity threshold parameter to control the degree of approximation.More details are given in Section5.4.5Visibility Culling with Hierarchical Occlusion MapsAn overview of the visibility culling algorithm has been presented in Section3.In this section,we present detailed algorithms for overlap tests with occlusion maps,depth comparison,and approx-imate culling.5.1Overlap Test with Occlusion MapsThe two-dimensional overlap test of a potential occludee against the union of occluders is performed by checking the opacity of the pixels it overlaps in the occlusion maps.An exact overlap test would require a scan-conversion of the potential occludee tofind out which pixels it touches,which is relatively expensive to do in software.Rather,we present a simple,efficient,and conservative solution for the overlap test.For each object in the viewing frustum,the algorithm conser-vatively approximates its projection with a screen-spacebounding rectangle of its bounding box.This rectangle covers a superset of the pixels covered by the actual object.The extremal values of the bounding rectangle are computed by projecting the cornersFigure4:Use of texture-mapping hardware to build occlusion mapsof the bounding box.The main advantage of using the bounding rectangle is the reduced cost offinding the pixels covered by a rectangle compared to scan-converting general polygons.The algorithm uses the occlusion map hierarchy to accelerate the overlap test.It begins the test at the level of the hierarchy where the size of a pixel in the occlusion map is approximately the same size as the bounding rectangle.The algorithm examines each pixel in this map that overlaps the bounding rectangle.If any of the overlapping pixels is not completely opaque1,the algorithm recursively descends from that pixel to the next level of the hierarchy and checks all of its sub-pixels that are covered by the bounding rectangle.If all the pixels checked are completely opaque,the algorithm concludes that the occludee’s projection is completely inside that of the occluders.If not,the algorithm conservatively concludes that the occludee may not be completely obscured by the occluders,and it is rendered.The algorithm supports conservative early termination in over-lap tests.If the opacity of a pixel in a low-resolution map is too low,there is small probability that we canfind high opacity values even if we descend into the sub-pixels.So the overlap test stops and concludes that the object is not occluded.The transparency thresholds are used to define these lower bounds on opacity below which traversal of the hierarchy is terminated.5.2Depth ComparisonOcclusion maps do not contain depth information.They provide a necessary condition for occlusion in terms of overlap tests in the image plane,but do not detect whether an object is in front of or behind the occluders.The algorithm manages depth information separately to complete the visibility test.In this section,we propose two algorithms for depth comparison.5.2.1A Single Z PlaneOne of the simplest ways to manage depth is to use a single plane.The Z plane is a plane parallel to and beyond the near plane. This plane separates the occluders from the potential occludees so that any object lying beyond the plane is farther away than any occluder.As a result,an object which is contained within the projection of the occluders and lies beyond the Z plane is com-pletely occluded.This is an extremely simple and conservative method which gives a rather coarse bound for the depth values of all occluders.5.2.2Depth Estimation BufferThe depth estimation buffer is a software buffer that provides a more general solution for conservatively estimating the depth of occluders.Rather than using a single plane to capture the depthViewerCD ABEFFigure 5:Distance criterion for dynamic selectionRedundancy:Some objects,e.g.a clock on the wall,pro-vide redundant occlusion and should be removed from thedatabase.Rendering Complexity:Objects with a high polygon count or rendering complexity are not preferred,as scan-converting them may take considerabletime and affect the overall frame rate.5.3.2Dynamic SelectionAt runtime,the algorithm selects a set of objects from the oc-cluder database.The algorithm uses a distance criterion,size,and temporal coherence to select occluders.The single -plane method for depth comparison,presented in Section 5.2.1,is also an occluder selection method.All objects not completely beyond the -plane are occluders.When the algorithm uses the depth estimation buffer,it dynam-ically selects occluders based on a distance criterion and a limit ()on the number of occluder polygons.These two variables may vary between frames as a function of the overall frame rate and percentage of model culled.Given ,the algorithm tries to find a set of good occluders whose total polygon count is less than .The algorithm considers each object in the occluder database lying in the viewing frustum.The distancebetween the viewer and the center of an object’s bounding volume is used as an estimate of the distance from the viewer to the object.The algorithm sorts these distances,and selects the nearest objects as occluder until their combined polygon count exceeds .This works well for most situations,except when a good occluder is relatively far away.One such situation has been shown in Figure 5.The distance criterion will select ,,,,etc.as occluders,but will probably be exceeded before and are selected.Thus,we lose occlusion that would have been contributed by and .In other words,there is a hole in the occlusion map which decreases the culling rate.Dynamic occluder selection can be assisted by visibility pre-processing of the occluder scene.The model space is subdivided by a uniform grid.Visibility is sampled at each grid point by surrounding the grid point with a cube and using an item buffer algorithm similar to the hemi-cube algorithm used in radiosity.Each grid point gets a lists of visible objects.At run-time,occlud-ers can be chosen from visible object lists of grid points nearest to the viewing point.5.4Approximate Visibility CullingA unique feature of our algorithm is to perform approximate vis-ibility culling,which ignores objects only visible through small holes in or among the occluders.This ability is based on an inherent property of HOM:it naturally represents the combined occluder projections at different levels of detail.In the process of filtering maps to build the hierarchy,a pixel in a low resolution map can obtain a high opacity value even if a small number of its descendant pixels have low opacity.Intuitively,a small group of low-opacity pixels (a "hole")in a high-resolution map can dissolve as the average operation (which involves high opacity values from neighboring pixels)is recursively applied to build lower-resolution maps.The opacity value above which the pixel is considered com-pletely opaque is called the opacity threshold ,which is by default 10.The visibility culling algorithm varies the degree of approx-imation by changing the opacity threshold.As the threshold is lowered,the culling algorithm becomes more approximate.This effect of the opacity threshold is based on the fact that if a pixel is considered completely opaque,the culling algorithm does not go into the descendant pixels for further opacity checking.If the opacity of a pixel in a low-resolution map is not 10(because some of the pixel’s descendents have low opacities),but is still higher than the opacity threshold assigned to that map,the culling algo-rithm does not descend to the sub-pixels to find the low opacities.In effect,some small holes in higher-resolution maps are ignored.The opacity threshold specifies the size of the holes that can be ignored;the higher the threshold,the smaller the ignorable holes.The opacity thresholds for each level of the hierarchy are com-puted by first deciding the maximum allowable size of a hole.For example,if the final image is 10241024and a map is 6464,then a pixel in the map corresponds to 1616pixels in the final image.If we consider 25black pixels among 1616total pixels an ignorable hole,then the opacity threshold for the map is 1251616090.Note that we are considering the worst case where the black pixels gather together to form the biggest hole,which is roughly a 55black block.One level up the map hierarchy,where resolution is 3232and where a map pixel corresponds to 3232screen pixels,the threshold becomes 1253232098.Consider the -th level of the hierarchy.Let black pixels among total pixels form an ignorable hole,then the opacitythreshold is14114Let the opacity threshold in the highest resolution map be .If a pixel in a lower resolution map has opacity lower than ,then it is not possible for all its descendant pixels have opacitiesgreater than.This means that if a high-level pixel is com-pletely covered by a bounding rectangle and its opacity is lowerthan,we can immediately conclude that the corresponding object is potentially visible.For pixels not completely covered by the rectangle (i.e.pixels intersecting the rectangle’s edges),the algorithm always descends into sub-pixels.To summarize the cases in the overlap test,a piece of pseudo-code is provided in 5.4.Approximate visibility is useful because in many cases we don’t expect to see many meaningful parts of the model through small holes in or among the occluders.Culling such portions of the model usually does not create noticeable visual artifacts.。
Cabibbo Allowed D - K pi gamma Decays
for the calculation of weak transition elements. We consider the use of this approach to be justified by the ”near” success of the approach for the nonleptonic amplitudes. This will involve its use in the D (D ∗ )Kπ vertices as well as in D (D ∗) → V and D (D ∗) → P transitions, all of which required for the calculations of the D → Kπγ amplitudes within our model. For the evaluation of the (D, D ∗ ) → (P, V ) transitions, we use the information obtained for these matrix elements from semileptonic decays (see, e.g. [20]). The general theoretical framework for our calculation is that of the heavy quark chiral Lagrangian [21, 22]. In the K → ππγ decays, it has been shown that intermediate light vector mesons play an important role in the decay amplitude [11]. We shall investigate the role of intermediate light vector mesons also in the D → Kπγ amplitude. In order to accomplish this, we use the extension of the formalism of [21, 22] to include also the light vector mesons [23, 24]. ¯ 0 π + γ and D 0 → K − π + γ shows that the direct part of The present study of D + → K the radiative amplitude is not much smaller in strength than the bremsstrahlung part, rather similarly to the inhibited K decays mentioned above. If confirmed by experiments, it places these decays in the status of a most suitable ground for the investigation of the mechanisms involved in such nonleptonic D decays. In Section II we present the theoretical framework for our calculation. In Section III we display the explicit expressions of all the calculated decay amplitudes. Section IV contains the discussion and the summary.
The multi-intersection signal control study
Key words: Multi-intersection Control, Delay, Genetic Algorithm, Simulation
Page 1
Contents
1. 2. INTRODUCTION .......................................................................................................... 2 ANALYSIS OF THE PROBLEM.................................................................................. 2 2.1 PHASE............................................................................................................................ 3 2.2 CYCLE TIME .................................................................................................................. 3 2.3 SIGNAL OFFSET ............................................................................................................. 3 3. 4. 5. TERMINOLOGY ........................................................................................................... 4 ASSUMPTIONS: ............................................................................................................ 4 MODEL AND RESULTS ............................................................................................... 5 5.1 THE ISOLATED INTERSECTION MODEL .......................................................................... 5
全双值格致死解法
全双值格致死解法全双值格致死解法是一种用于解决死循环问题的方法，被广泛运用于计算机科学领域中的各种算法和数据结构。
在本文中，我们将详细介绍全双值格致死解法的定义、原理和应用，以及解决死循环问题的其他方法和技巧。
一、全双值格致死解法的定义和原理1.1定义全双值格（Bivalent Grid）是指一个二维数据结构，每个格子只能存储两种状态的值：0或1。
在全双值格的基础上，全双值格致死解法提出了一个新概念：“致死区域”（Dead Zone）。
如果一个全双值格中的某个区域被设置成了致死区域，那么在处理这个区域时，算法就会陷入死循环。
1.2原理全双值格致死解法的原理是，通过在算法中加入全双值格数据结构和致死区域的定义，可以有效地避免死循环的发生。
具体而言，算法在执行过程中，先将全双值格中所有的格子初始化为0，然后根据问题的特点，将某些格子设为1，并将它们所在的区域标记为致死区域。
在算法中的每个循环迭代中，检查算法所处理的数据是否属于致死区域，如果是，则立即停止算法的迭代，返回错误结果。
二、全双值格致死解法的应用全双值格致死解法可以应用于多种算法和数据结构中，例如迷宫问题、图论问题、自动机、计算几何等。
下面分别以迷宫问题和图论问题为例，说明全双值格致死解法的应用。
2.1迷宫问题迷宫问题是指在一个二维格子中寻找从起点到终点的路径。
在一般的迷宫问题中，某些格子是墙，无法通过；而其他格子可以通过。
遍历整个迷宫找到一条从起点到终点的路线是一个经典的问题。
在解决迷宫问题时，可以应用全双值格致死解法来防止算法运行进入死循环。
以求解迷宫问题的回溯算法为例，进行说明。
回溯法是一种在问题的所有可能解空间搜索过程中，每到达一个状态时，先自上而下依次访问每个子节点，直到达到不能继续向下搜索的节点为止，然后就回溯到上一节点，继续向下搜索其它的子节点。
在求解迷宫问题时，我们可以将该问题的解空间分为四个方向 :上、下、左、右。
在搜索与回溯的过程中，我们可以利用全双值格数据结构和致死区域的定义，来避免算法运行进入死循环。
- 1、下载文档前请自行甄别文档内容的完整性，平台不提供额外的编辑、内容补充、找答案等附加服务。
- 2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
- 3、如文档侵犯您的权益，请联系客服反馈,我们会尽快为您处理(人工客服工作时间：9:00-18:30)。
Calculation of Deadline Missing Probability in a QoS Capable ClusterInterconnectEun Jung Kim Ki Hwan Yum Chita R.DasDepartment of Computer Science and EngineeringThe Pennsylvania State UniversityUniversity Park,PA16802E-mail:ejkim,yum,das@AbstractThe growing use of clusters in diverse applications,many of which have real-time constraints,requires Quality-of-Service(QoS)support from the underlying cluster inter-connect.In this paper,we propose an analytical model thatcaptures the characteristics of a QoS capable wormholerouter,which is the basic building block of cluster networks.The model captures the behavior of integrated traffic in acluster and computes the average deadline missing proba-bility for real-time traffic.The cluster interconnect,consid-ered here,is a hypercube parison of DeadlineMissing Probability(DMP)using the proposed model withthat of the simulation shows that our analytical model isaccurate and useful.Index Terms:Analytical Model,Cluster Network,SAN,Quality-of-Service,Pipelined Router Architecture,Virtual-Clock,Wormhole Switching.1IntroductionQuality-of-Service(QoS)provisioning in clusters hasbecome a critical issue with the widespread use of clus-ters in diverse commercial applications.The traditionalbest-effort service model that has been used for scientificcomputing is not adequate to support many cluster appli-cations with varying consumer expectations.For example,bandwidth proportionately[19,24].Techniques such as preemption of lower priority traffic in favor of higher prior-ity traffic have also been proposed[21].Recently,we have proposed a QoS-aware pipelined router that supports fea-tures such as rate-based scheduling,preemption,andflit ac-celeration mechanism[23].Software solution like the self-synchronizing scheduling[5]does not need any hardware modification,but the solution may not be scalable.A limitation of all prior studies is that they use simulation to evaluate the performance of various design trade-offs.In addition,the evaluations are confined to a single router in many cases.Detailedflit-level simulation is quite expensive and prohibits full-blown analyses of various design trade-offs.On the other hand,an accurate analytical model can provide quick performance estimates and will be a valuable design tool.In this paper we present a mathematical model for analyzing QoS capable cluster networks.In[15],we had developed a mathematical model of a QoS-aware clus-ter network to compute the average network latency.While the average latency is an important performance metric for all types of traffic,it does not capture the behavior of real-time traffic in sufficient detail.For example,if an applica-tion messages requires low jitter tolerance,then jitter will be a main performance metric.In our case,since we consider time-constraint applications,delay bound is the primary ob-jective function.Since wormhole-switched network cannot provide hard guarantees due to chained blocking,the sys-tem can provide soft guarantees in terms of deadline miss-ing probability(DMP).The DMP of time-constrained ap-plications was analyzed in[19,14]via simulation.In this paper,we present an analytic approach to compute DMP of real-time applications.Wefirst develop the model for a single router and then extend it to a network.Here we use a hypercube-style clus-ter network primarily to keep the analysis tractable due to the symmetric nature of the network.However,our QoS-aware router model can be extended to any regular topology such as-ary-cubes and meshes as long as the topology and routing algorithm can be captured mathematically.Like many commercial designs,we use a pipelined wormhole router architecture.The model considers an inte-grated workload consisting of different classes of traffic.classes represent real-time applications1with dis-tinct service requirements.The last class is used for best-a full crossbar since it has inputs and out-puts.The model can be modified for a multiplexed crossbar, where the VC multiplexing will be done before the crossbar stage.n-1Figure 1.The pipelined router architecturewith a full crossbar.Unlike the lumped router models analyzed before[1,16, 6,13]that capture only the blocking delay caused by arbi-tration,a message entering the above pipelined router can experience delay at stages1,3and5of the router.If the corresponding input buffer is full in stage1,the message must wait outside the router until adequate space is avail-able.In stage3,the message again may be delayed because its destination crossbar output port could be busy.Crossbar output port arbitration is performed at a message level gran-ularity.So the message has to wait until the output port is re-leased by the message currently using it.Finally in stage5, multiple VCs compete for the physical channel bandwidth. Traditionally,a Round Robin or FIFO scheduler is used to schedule the output channel in a time-division manner.The above router design is modified to support QoS pro-visioning by simply incorporating a rate-based scheduling algorithm to share the physical channel bandwidth.Similar techniques have been proposed for the Internet router line cards.Our previous research[24,23,15]has shown that this architecture with the VirtualClock algorithm[25]can be effective in providing QoS for integrated traffic.In the VirtualClock algorithm,there are two variables, called auxVC and Vtick for each connection.The values of these two variables are determined when a connection is set up.The auxVC indicates the virtual clock value of the con-nection,while the Vtick is the amount of time that should be incremented whenever aflit arrives at that connection.The Vtick value specifies the interarrival time offlits from the connection.Therefore,a smaller Vtick value implies higher bandwidth.Once these two values are set,the VirtualClock algorithm works as follows.For each connection,when aflit arrives at the scheduler,the following computation is done.auxVC real time auxVCauxVC auxVC Vticktimestamp theflits with the auxVCTheflits are queued and serviced in increasing timestamp order.For the best-effort traffic,the timestamp is set as .So the best-effortflits are processed only if there are no otherflits with lower timestamp values.3Deadline Missing ProbabilityAs described in the previous section,the router model assumes a pipelined architecture with stages.The model is derived for classes of traffic with different ser-vice requirements.Here we assume that there arereal-time traffic classes and one class of best-effort traffic. Each class is assigned a dedicated VC.(This assumption can be relaxed to assign multiple VCs to a class.)In addi-tion,the model is based on the following assumptions typi-cally used in analytical models:The arrival pattern of each class follows the Poisson processes with an average arrival rate of.Message length isflits long.Message destination is uniformly distributed.The Vtick value for real-time traffic belonging to classis given by,and the Vtick value for best-effort traffic is set to.The input and output buffers(VCs)in stages1and5 can holdflits.Each class is assigned a dedicated injection/ejection queue outside the router,and these queues have infinite capacity.For a given source and destination pair,the probability of missing the deadline is the probability that a message cannot be delivered within a specified time().We only compute the deadline missing probability of real-time traffic which has time-constraints.Here we consider only deadline to traverse the network.Source queueing is not included in order to keep the discussion simple.3.1Single Router ModelWe compute the DMP for a class traffic in a single router.The network latency of class,,is the time to traverse the router.The network latency()of a message of class consists of two parts.Thefirst part is the actual message transfer time,.The second part is due to block-ing caused by the wormhole switching scheme,and due to sharing of the physical channel bandwidth by multiple vir-tual channels at stage5of Fig.1.The actual transmission time with pipeline stages in a single router iscycles for an-flit message.In order to compute the second part of the network la-tency,let us define as the blocking length(in number of flits)seen by the headerflit at the input,output,and arbitra-tion stage in the router.captures the message blocking in a pipelined wormhole router.Then the effective length of the message becomesflits.Let be the aver-age number of cycles required to transfer oneflit of a class message.represents the effect of bandwidth sharing mechanism of the Virtual Clock algorithm.Thus,the net-work latency()for is(1)While blocking happens among the same class of mes-sages,the sharing depends on the traffic of other classes. Thus,these two random variables(and)are inde-pendent.We can combine them to a random parameter,.We know thatand.Let be the probability of missing the deadline .If we canfind the c.d.f.of,then is),where.For accurate estimation of,first we consider the two random variables(and)separately and then combine them.To compute the blocking length,note that block-ing is possible at the input buffer stage,output buffer stage and arbitration stage.The worst case of blocking occurs when all these places are occupied by other messages.Thus the worst blocking length will bewhere is the input/output buffer size and M is the mes-sage length.The term is used to capture the buffer length,since a new message must wait until the service for the previous message is completed. Let us assume that we know the probability mass functionof(),which will be described later.With a given blocking delay(),the effective message length will be.When eachflit of arrives at the head of the output VC,there are combinations of other real-time traffic that denote whether they occupy the corresponding output VCs or not.All these combinations will determine how to share the bandwidth.We number the combinations serially so that for each combination()we can determine the number of cycles required to transfer aflit of class traffic,,and the probability of th combination for traffic,.Let be the number offlits,which needs cy-cles at the output VC such that,given that the blocking length is.Then,the actual de-lay for a blocking length can be denoted asLet’s define the c.d.f of,,as(2) There are summation notations in Eq.2.Thefirst notation is for and the remaining notations correspond to the total number of combinations of the out-put VC status.In Eq.2,is the probability that given the blocking length is.,the upper bound of,is which is the worst case of blocking,,for.We need the solution of and to find the deadline missing probability.Since the exact es-timation of the terms is extremely hard,we approximate these probabilities from the operational behavior of the router/network.If we have the blocking probability of class 2,,then.Since the blocking length()varies between0and,underthe uniform distribution assumption,can be writ-ten as(3)otherwise,where.Similarly we can get with,whichagain for better readability3,is deferred to the Appendix.Since,and.From Eq.3,can be written asotherwise(8)where is the blocking probability of class in channel .Similarly from Eq.4,we can get with asotherwise.Note that all these equations can be derived from the sin-gle router model for a given number of hops()and for a physical channel by setting the proper boundary values.4Performance ResultsUsing the equations derived in Section3,we compute the DMPs in a single router and in a6-cube.Some of the results are presented here for validating the model.We are unable to include results of other cube dimensions due to space limitation.We also implemented a corresponding simulation model as shown in Fig.1using CSIM.Note that we need a deadline parameter to estimate the DMP.In our pipelined router model,the minimum transfer time for a32-flit message is36cycles().Hence,we set cycles for the single router.Similarly for2-hop messages in a6-cube,the minimum transfer time is46cy-cles().We set or60cycles for 2-hop messages.In Fig.2,we plot the DMP results for two types of real-time traffic(R1and R2)from the mathematical model (Math)and the simulation model(Sim).In a6-cube,the DMPs of2-hop and5-hop messages are shown for different values.The graphs show that the single router results are more accurate compared to the6-cube results.This is be-cause we approximate the upper bound of blocking length in each hop to max without accounting for the chained blocking.Since there is no chained blocking in a single router,the upper bound approximation is more accurate.Even with this approximation,the DMP results from the analytical model of a6-cube match closely with the simulation results.5Concluding RemarksThis paper introduces an analytical approach for calcu-lating the DMP of real-time traffic in a QoS-aware worm-hole router and a hypercube-style cluster network,designed using such routers.For accurate calculation,the model cap-tures the pipelined design,and analyzes the blocking delay at different stages of the pipe.In addition,the effect of Vir-tualClock scheduling algorithm is reflected in the model. Comparison with the simulation results indicates that the router as well as the hypercube models are quite accurate in predicting the DMP.Unlike the simulation model,the analytical model can be used as an efficient design tool in studying various design trade-offs.For example,the im-pact of message length(),and other questions can be an-swered quickly using the model either for a single-cluster or for a multi-router cluster.Such performance estimates and quick design overviews are difficult to obtain via a simula-tion study.The model presented here can be improved in a variety of ways,and some of them are currently pursued in our group. First,the exponential arrival distribution for real-time traf-fic may not be quite practical to apply to media streams.We need to develop the model with a CBR/VBR source to cap-ture inputs like media streams.Second,QoS comes with different connotations,and extension of the model to pre-dict other performance parameters such as bandwidth as-surance and jitter should be useful.Third,the model can be extended to other topologies.Finally,co-evaluation of the cluster network with a detailed network interface model should answer many questions regarding the QoS ability of the entire communication system.References[1]V.S.Adve and M.K.Vernon.Performance Analysis ofMesh Interconnection Networks with Deterministic Rout-ing.IEEE Transactions on Parallel and Distributed Systems, 5(3):225–246,March1994.[2]N.J.Boden,D.Cohen,R.E.Felderman,A.E.Kulawik,C.L.Seitz,J.N.Seizovic,and W.-K.Su.Myrinet:AGigabit-per-second Local Area Network.IEEE Micro, 15(1):29–36,February1995.[3]M.B.Caminero,J.J.Quiles,J.Duato,D.S.Love,andS.Yalamanchili.Performance Evaluation of the Multimedia Router with MPEG-2Video Traffic.In Proceedings of the Third International Workshop on Communication,Architec-ture and Applications on Network Based Parallel Computing (CANPC’99),pages62–76,January1999.Real−time Load (messages/cycle)0.000.050.100.150.20D e a d l i n e M i s s i n g P r o b a b i l i t y(a)Single router with deadline 42cyclesReal−time Load (messages/cycle)0.000.100.200.300.400.500.60D e a d l i n e M i s s i n g P r o b a b i l i t yReal−time Load (messages/cycle)0.000.100.200.300.400.500.60D e a d l i n e M i s s i n g P r o b a b i l i t y(b)Deadline :55cycles (2hops),70cycles (5hops)(c)Deadline :60cycles (2hops),75cycles (5hops)Figure 2.DMP comparison of analytical model and simulation model in a single router and 6-cube with varying real-time load and fixed best-effort load (single router :0.01msgs/cycle,6-cube :0.002msgs/cycle).[4]J.Carbonaro and F.Verhoorn.Cavallino:The TeraflopsRouter and NIC.In Proc.Symp.High Performance Intercon-nects (Hot Interconnects 4),pages 157–160,August 1996.[5]K.Connelly and A.A.Chien.FM-QoS:Real-Time Commu-nication Using Self-Synchronizing Schedules.In Proceed-ings of Supercomputing Conference ,November 1997.[6]W.J.Dally.Performance Analysis of -ary -cube Inter-connection Networks.IEEE Transactions on Computers ,39(6):775–785,June 1990.[7] A.Demars and S.Shenker.Analysis and Simulation ofa Fair Queueing Algorithm.In Proceedings of the ACM SIGCOMM Conference on Applications,Technologies,Ar-chitectures,and Protocols for Computer Communication ,pages 1–12,1989.[8]J.T.Draper and J.Ghosh.A Comprehensive AnalysisModel for Wormhole Routing in Multicomputer Systems.Journal of Parallel and Distributed Computing ,32:202–214,1994.[9]J.Duato,S.Yalamanchili,M.B.Caminero,D.Love,andF.J.Quiles.MMR:A High-Performance Multimedia Router-Architecture and Design-Tradeoffs.In Proceedings of the IEEE International Symposium on High-Performance Computer Architecture ,pages 300–309,January 1999.[10]H.Eberle and E.Oertli.Switcherland:A QoS Communica-tion Architecture for Workstation Clusters.In Proceedings of the International Symposium on Computer Architecture ,pages 98–108,June 1998.[11]M.Galles.Scalable Pipelined Interconnect for DistributedEndpoint Routing :The SGI SPIDER Chip.In Proceed-ings of Symposium on High Performance Interconnects (Hot Interconnects),pages 141–146,August 1996.[12] D.Garcia and W.Watson.Servernet II.In Proceedings ofthe1997Parallel Computing,Routing,and Communication Workshop(PCRCW’97),June1997.[13]P.T.Gaughan and S.Yalamanchili.A Performance Model ofPipelined-ary-cubes.IEEE Transactions on Computers, 44(8):1059–1063,Auguest1995.[14] B.Kim,J.Kim,S.Hong,and S.Lee.A Real-Time Commu-nication Method for Wormhole Switching Networks.In Pro-ceedings of International Conference on Paralle Processing, pages527–534,August1998.[15] E.J.Kim,K.H.Yum,and C.R.Das.An AnalyticalModel for a QoS Capable Cluster Interconnect.To be pre-sented at11th GI/ITG Conference on Measuring,Modelling and Evaluation of Computer and Communication Systems (MMB2001),September2001.[16]J.Kim and C.R.Das.Hypercube Communication Delaywith Wormhole Routing.IEEE Transactions on Computers, 43(7):806–814,July1994.[17]J.H.Kim.Bandwidth and Latency Guarantees in Low-Cost,High-Performance Networks.PhD thesis,Department of Electrical Engineering,University of Illinois at Urbana-Champaign,1997.[18]J.H.Kim and A.A.Chien.Rotating Combined Queue-ing(RCQ):Bandwidth and Latency Gurantees in Low-Cost, High-Performance Networks.In Proceedings of the Inter-national Symposium on Computer Architecture,pages226–236,May1996.[19]J.-P.Li and M.Mutka.Priority Based Real-Time Communi-cation for Large Scale Wormhole Networks.In Proceedings of International Parallel Processing Symposium,pages433–438,May1994.[20]S.L.Scott and G.M.Thorson.The Cray T3E Network:Adaptive Routing in a High Performance3D Torus.In Pro-ceedings of Symposium on High Performance Interconnects (Hot Interconnects),pages147–156,August1996.[21]H.Song, B.Kwon,and H.Yoon.Throttle and Pre-empt:A New Flow Control for Real-Time Communications in Wormhole Networks.In Proceedings of International Conference on Paralle Processing,pages198–202,August 1997.[22] C.B.Stunkel,D.G.Shea,B.Abali,M.G.Atkins,C.A.Bender,D.G.Grice,P.Hochschild,D.J.Joseph,B.J.Nathanson,R.A.Swetz,R.F.Stucke,M.Tsao,and P.R.Varker.The SP2High-Performance Switch.IBM Systems Journal,34(2):185–204,1995.[23]K.H.Yum,E.J.Kim,and C.R.Das.QoS Provisioning inClusters:An Investigation of Router and NIC Design.In Proceedings of the International Symposium on Computer Architecture,pages120–129,June2001.[24]K.H.Yum,A.S.Vaidya,C.R.Das,and A.Sivasubrama-niam.Investigating QoS Support for Traffic Mixes with the MediaWorm Router.In Proceedings of the IEEE Interna-tional Symposium on High-Performance Computer Archi-tecture,pages97–106,January2000.[25]L.Zhang.VirtualClock:A New Traffic Control Algorithmfor Packet-Switched Networks.ACM Transactions on Com-puter Systems,9(2):101–124,May1991.AppendixComputation of blocking probability()in a single router:Here we describe the computation of blocking probabil-ity in a single router.Since the input/output buffer sizes are ,the blocking probability,for class(), can be expressed as(9)where is the steady state message arrival rate of class traffic.(are considered at the message-level gran-ularity,the total buffer size(flits)of input and output queues becomes when converted to message length. (Note that we are considering the worst case scenario here by using the entire buffer length.)Including the cur-rently servicedflit/message,the total number of messages becomes.Hence,the channel utilization(or blocking probability of class)is given by Eq.9.The steady state arrival rate in Eq.9is given by(10)The average network latency(,(11)where.Note that due to the inter-dependencies between and,the solution becomes iterative.Computation of blocking probability()in a Clus-ter Interconnect:In order to get unknown term in Eq.8,we need to com-pute the blocking probability()in a Cluster Intercon-nect.From Eq.9,the probability of blocking for class traffic in channel can be written asbe the total transit message rate of traffic at a router.The generation rate of traffic in the steady state is.There-fore,the total message rate at the output of a router(over all the output channels)is.Let be the transit message arrival rate of traffic from other nodes at physical channel of a router.Similarly,represents traffic generated by the source node for a physical channel and virtual channel.We give here the expressions derived in[16]for completeness.The transient message arrival rate at physical channel and virtual channel of a router is given bySimilarly Eq.10is modified as(13)In Eq.12,is the latency of a class message when it uses in physical channel as thefirst path to traverse to-wards its destination.can be expressed as(14) ()in Eq.14de-notes the average number of hops a message travels starting with the physical channel as thefirst path.()is the blocking length of a class real-time message at stage1of thefirst router that uses channel as thefirst route,and() is the blocking length at stages3and5in the ejection chan-nel of the last router.Also,middle is the block-ing length between the source and the destination(i. e. middle nodes)excluding the blocking length at stage1of the source and the blocking length at stages3and5of the destination.The computation of middle will be described later.is the average number of cycles perflit in the ejection channel.Computation of the probability of output VCs status, :The probability of th combination for class,, can be determined using a Markov model.Let be a state such that the th output buffer is empty and be the state such that the th output buffer is not empty.The status of the rest buffers are all identical in the two states to make and adjacent.Let’s assume the serial number of on class be.(The detailed numbering function can be found in[15].)Now,the transition rate from state to is,where is the traffic rate of the th VC(Eq.10), while the rate from to is,whereVtickwhere th VC is busy(15) where the serial number of on class is k.The probability of output VCs status()in the net-work can be obtained similarly.Detailed computation of these probabilities can be found in[15]. Computation of middle for Eq.14:To compute middle,we use the delay model from [16],except that we include the input and output queue-ing delay,while in[16]they capture only blocking delay. The average length of blocking in the middle nodes for a message which uses physical channel as thefirst path, middle,ismiddleis the probability that a message does not terminate after using physical channel as thefirst path,and the last fractional expression represents the average number of hops the message travels when it takes a channel after using.The average length of messages involved in blocking for each channel is given aswhere。