Traffic engineering with constraint based routing in MPLS networks

Traffic Engineering with Constraint Based Routing in MPLS NetworksHaris Hodzic, Sladjana ZoricBH Telecom, Zmaja od Bosne 88, 71000 Sarajevo, Bosnia and HercegovinaE-mail: haris.hodzic@bhtelecom.baAbstract- In today’s service provider networks MPLS Traffic Engineering (MPLS TE) is one of th e fastest growing i mplementation. MPLS adoption in service provider networks h as increased manifold due to its inherent TE capabilities. A purpose of traffic engineering in a network is to setting up paths between the edge nodes of th e network to meet traffic demands wh ile optimizing network performance. It is known th at total traffic throughput in a network, or resource utilization, can be maximized if a traffic demand is split over multiple path s. MPLS networks can use native TE mech anisms to minimize network congestion and improve network performance. TE modifies routing patterns to provide efficient mapping of traffic streams to network resources. Th is efficient mapping can reduce th e occurrence of congestion and can play an important role in th e implementation of network services with quality of service (QoS) guarantees. Th ese MPLS TE capabilities bring explicit routing, constraint-based routing, and bandwidth reservation to MPLS networks. MPLS TE relies on extensions to existing IP protocols (TE-RSVP, CR-LDP).Keywords – MPLS, Traffic Engineering, LSP, Constraint Based Routing, CR-LDP, RSVP-TE1.INTRODUCTIONTE is the mechanism used for steering traffic across to the backbone to facilitate efficient use of available bandwidth between a pair of routers. Prior to MPL S TE, traffic engineering was performed either by IP or by ATM, depending on the protocol in use between two edge routers in a network.There are four basic components needed for traffic engineering in a packet network:• Distribution of topology information: Because the networkis a distributed system, it is necessary to have a mechanism to advertise up-to-date information about links to all nodes in the network. A node can then build a topology map of the network. Information about link or node failures must be rapidly propagated through the network so that recovery can be initiated.• Path selection: This involves using the topology information gathered to compute reachability information between nodes in the network. When a path is required between any pair of nodes in the network, a shortest path with a minimum of links is typically used because it optimizes the use of resources.Other constraints, such as bandwidth and delay, may also be applied during path selection.• Directing traffic along the computed paths: Once the path between a given source node and destination node has been computed, we need to be able to direct traffic along that particular path. Typically, this is accomplished by building a forwarding table. When a packet arrives at a node, the forwarding table is consulted using information in the packet header. In a connectionless forwarding paradigm, each node builds its forwarding tables independently.For connection-oriented forwarding this step involves use of a signaling protocol to set up the path. The signaling protocol assists intermediate nodes along the path in building their forwarding tables.• Traffic management: Traffic management deals with the definition of a framework and the corresponding mechanisms that enable a network to deliver predictable QoS to user traffic. QoS is characterized by parameters such as bandwidth, delay, delay jitter, and loss probability. The mechanisms for implementing a QoS solution include admission control, flow identification, traffic policing, and traffic scheduling [1].In this paper we represent overview of MPLS TE in chapter 2, and Constraint Based Routing with comparing two most utilized signaling protocols in chapter 3. In chapter 4 simple example of MPL S network is simulated with and without using of traffic engineering mechanisms.2.MPLS TEIn a traditional IP forwarding paradigm, packets are forwarded on destination basis where a route lookup is performed on each router from source to destination. Destination-based forwarding paradigm leads to suboptimal use of available bandwidth between a pair of routers in the service provider network. Predominantly, the suboptimal paths are under-utilized in IP networks. To avoid packet drops due to inefficient use of available bandwidth and toprovide better performance, TE is employed to steer some of the traffic destined to follow the optimal path to a suboptimal path to enable better bandwidth management and utilization between a pair of routers. TE, hence, relieves temporary congestion in the core of the network on the primary or optimal cost links. TE maps flows between two routers appropriately to enable efficient use of already available bandwidth in the core of the network. The key to implementing a scalable and efficient TE methodology in the core of the network is to gather information on the traffic patterns as they traverse the core of the network so that bandwidth guarantees can be established. Main mission of MPLS TE is to maximize network resources [2].The inspiration behind MPL S TE is Constraint Based Routing (CBR), which takes into account the possibility of multiple paths between a specific source/destination pair in a network. With CBR, the operation of an IP network is enhanced so the least cost routing can be implemented as well as variables to find paths from a source to destination. CBR isthe backbone of the TE tunnel definition and isdefined on the ingress routers to the MPLS domain when implementing MP L S TE. Resource availability and link status information are calculated using a constrained SPF calculation in which factors such as the bandwidth, policies, and topology are taken into consideration to define probable paths from a source to destination. CSPF calculation results with an ordered set of IP addresses that map to next-hop IP addresses of routers forming an LSP, in turn mapping to the TE tunnel. This ordered set is defined by the headend router that is propagated to other routers in the LSP. The intermediate routers, thus, do not perform the function of path selection. RSVP and L DP (L abel Distribution Protocol) with TE extensions are used to reserve resources in the LSP path as well as label association to the TE tunnel [3].3.CONSTRAINT BASED ROUTINGThere are two ways to set up a label switch pathwithin a MPLS network, control driven or explicitly routed. Control driven LSP can be set up by hop-by-hop routing or L DP, which involves setting up a connection thru UDP and TCP. Second approach is ER-L SP. An Explicit route is a small sequence of hops from ingress to egress LSRs to set up an LSP. This explicit route can contain several hops within the set of many nodes, within an MPLS environment before emerging to the next hop specified in the Explicit Route. Explicit routing helps in diversion of network traffic around failed links and helps inproviding already set up L SP-backup to maintain uninterrupted flow. Explicit routing has its significance to force an LSP, which differs from the one offered by the routing protocol. Constraint-based Routing computes routes that are subject to constraints such as bandwidth and administrative policy. Using a combination of the metrics defined for traffic engineering and the capabilities of routers, constraint-based routing substantially reduces the requirements for operator activity necessary to implement TE. Because Constraint-based Routing considers more than network topology in computing routes, it may find a longer but lightly loaded path better than the heavily loaded shortest path. Network traffic is hence distributed more evenly. Traffic Engineering RSVP or TE-RSVP and Constrained-Based L DP or CR-L DP developed by the IETF MPLS Working Group follows setting ER-LSP [4].4.MPLS TE SIGNALING PROTOCOLS In an MPLS network an LSP must be set up and labels assigned at each hop before traffic forwarding can take place. As we said earlier there are two kinds of L SPs based on the method used for determining the route: control driven L SPs (also called hop by hop LSPs), and explicitly routed LSPs (also referred to as constraint based routed LSPs, CR-LSPs). When setting up a control-driven L SP, each LSR determines the next interface for the L SP based on its IP forwarding table, and sends the label request to the next-hop router. When setting up a CR-LSP, the route for the LSP is specified in the setup message. The setup message traverses all nodes along the specified route. At each node, a label request is sent to the next indicated interface. Thus, a control driven LSP follows the path that a packet using default IP routing would have used. On the other hand, a CR-LSP may be specified and controlled by the network operator or a network management application to direct the network traffic on a path independent of what is computed by IP forwarding. In this way CR-L SPs may be used for traffic engineering. IETF MPLS brought up two signaling protocol called the L abel Distribution Protocols (L DP) for setup and maintenance of control-driven LSPs. For setting up CR-L SPs, two approaches are being discussed: constraint-based routed L DP (CR-L DP), which requires a subset of L DP functionality that is enhanced to signal explicit paths, and extensions to the RSVP protocol. Before compare at glance RSVP and CR-L DP protocols, it is worthwhile to look at the desirable features of a signaling protocol for such applications [5].270The requirements for a signaling protocol used in MPLS based traffic engineering are as follows:• Robustness: The signaling system must be able to ensure reliable and timely delivery of signaling messages even in the presence of network congestion or failure.• Scalability: The size of ISP networks requires support for a large number of LSPs at each node.• Specification of QoS: This includes the specification of the traffic descriptors (i.e., bandwidth requirements) associated with the traffic flow using an LSP, and the QoS requirements (e.g., delay, loss).• L SP establishment/teardown/maintenance: The signaling protocol must be able to provide L SP establishment, teardown, and maintenance.• L SP priority/preemption: Path preemption is a traffic engineering requirement to ensure that high-priority L SPs may preempt (i.e., tear down) lower-priority L SPs when there are not enough resources available to support both. • Flexibility in path setup options: This includes strict and loose CR-L SPs, as well as the option to pin loose segments of a path [6].L DP/CR-L DP offers a unified signaling protocolsystem that provides network operators with thecomplete label distribution and path setup modesneeded for MPLS. The extensions to RSVP providethe capability to establish CR-LSPs withdownstream-on-demand label allocation, distribution, and binding only. If other MPLS labeldistribution modes are needed in the network, suchas the downstream-unsolicited method of labeldistribution, the network will be required to run bothLDP and RSVP.CR-LDP uses TCP, ensuring reliable transport of its signaling messages. When a failure occurs, an error indication message is sent using reliable transport, ensuring a timely response. Due to the lack of a reliable transport infrastructure, RSVP cannot guarantee fast failure notification to the endpoints affected by the failure, even though an explicit teardown message is sent. CR-L DP is a hard-state protocol; because of this it provides better scaling properties in terms of the volume of signaling traffic in the network as the number of CR-LSPs increases. A significant drawback of RSVP is its scalability problem when there are a large number of paths passing through a node. This is due to the requirement for periodically refreshing the state for each path [7].5.MPLS TE PERFORMANECEFor simulation network utilizations process with (and without) using MPLS TE, we created topology show below using NS-2. Nodes 0, 1, 7 and 8 can be considered as computer or non-MPL S supporting device. Nodes 0 and 1act as sources of traffic, while nodes 7 and 8 are their respective sinks. In other words, nodes 1 and 0 send some data to nodes 8 and 7 through the MPLS network. Nodes 2, 4, 5, 6 and 9 act as MPL S enabled nodes. When no traffic engineering is done the packets get lost due to congestion at node 2, as both the traffic flow take the shortest path and it gets congested.There are 2 paths between ingress-egress, but only the path with minimum hop count is used (The path through node 2, 9 and 6). The other path (Path via node 2, 4, 5 and 6) is not utilized, while the shortest path gets over utilized.Figure 2. shows simulations results where we can see throughput of each flow when no traffic engineering is applied and all the traffic flow via the shortest path.In the other case we have same topology but now traffic engineering is applied. The point of congestion is identified and rectified, by switching one of the flows (Either flow from 0 to 7 or 1 to 8) through the explicit path via nodes 2,4,5 and 6. When this is there is no packet loss due tocongestion and the throughput of the network is increased. This scenario is depicted in figure 4. The delay in the case when traffic engineering is applied is lower then when it is not applied (ReferFigure 2 and Figure 4). Fig. 1. Packet loss due to congestionFig. 2. Graph showing the delay for each flow when the shortestpath algorithm is usedFig. 3. No congestion – After applying traffic engineering Fig. 4 Graph showing the delay for each flow when trafficengineering is applied6.CONCLUSIONMPLS TE mechanism has been addressed and the protocols were reviewed. Almost similar functionality has been accomplished in the establishment of traffic-engineered paths, using both the protocols. Each one of the protocols has its strengths and weaknesses. There has been a need foran extensive enhancement in a protocol that supports the functioning of MPL S more efficiently. As both CR–LDP and TE–RSVP evolve they will offer more and more similar functionality. As with all new protocols there is still a fair amount of work to be accomplished. Simulation results showed importance of implementing MPLS traffic engineering in a large networks, as well it’s contribution in optimizing network resources and maximizing network bandwidth.REFERENCES[1] A. Ghanwani, B. Jamoussi, D. Fedyk,“Traffic Engineering Standards in IPNetworks Using MPL S”, IEEECommunicationMagazine1999[2] “MPLS Traffic Engineering”, Cisco Press,2006[3] D. Awduche, J. Malcolm, J. Agogbua, M.O'Dell, J. McManus “Traffic Engineeringin MPLS”, RFC 2702.[4] V. Jolly, S. Latifi, “An Overview of MPLSand Constraint Based Routing ”[5] A. Singh, G. Mittal “QoS and TrafficEngineering: MPLS, DiffServ and Constraint Based Routing”, May, 2000.[6]W. Sun “QoS/Policy/Contraint Based Routing”,January,1999.[7] K. L ong, Z. Zhang, S. Cheng, “L oadbalancing algorithms in MP S trafficengineering”, High Performance Switchingand Routing, IEEE Workshop, 2001.272。