opnet 四天培训教程 Modeler_Day2

Beijing Credit Top Company Co., Ltd.北京怡嘉行科技有限公司OPNET Modeler Training (Day 2)2Beijing Credit Top Co. ,Ltd. All rights reservedCourse Content•DAY 1:•Introduction•Modeling Framework•Features of the Project Editor {Lab}•Node Modeling {Lab}•Process Modeling {Lab}•Other Editors {Lab}•Running The DES•Understanding Statistics •Configure/Run DES •Simulation Log•Viewing Simulation Results •Publishing and Animation •Building a First Network {Lab}•DAY 3:•Traffic Modeling •Modeling Applications with the Standard Application Models{Labs}•Debugging Techniques {Labs}DAY 2:•General Simulation Methodology•Events and Event List Concepts •Process ModelingMethodology (PMM) {Lab}•Parametric Studies{Lab}DAY 4:•Wireless Modeling {Labs}•Modeling Custom WirelessEffects {Labs}Beijing Credit Top Company Co., Ltd.北京怡嘉行科技有限公司General Simulation Methodology4Beijing Credit Top Co. ,Ltd. All rights reserved•Topics–Understanding what you are modeling –Granularity –Defining Data –Results•Conceptual Goals–Use basic methodology for developing a network simulation study.–Plan simulation studies using a consistent method.Agenda5Beijing Credit Top Co. ,Ltd. All rights reservedFlow Diagram: A Simulation StudyYesEnd StartResultsstatisticallyuseful?NoNoResultssufficientlydetailed?Choosing input andrunning simulationsSystem resultsaccurate?Defining input andoutputSpecifying thesystem modelNoChoosing aspectsto be modeledUnderstanding yourgoals for thesimulationUnderstanding thesystem6Beijing Credit Top Co. ,Ltd. All rights reservedUnderstanding the Proposed System•You must understand the proposed system to model it accurately.•The model cannot be more accurate than your understanding.•Recommendation: obtain a written specification of the system you will be modeling.7Beijing Credit Top Co. ,Ltd. All rights reservedUnderstanding Your Goals•What general questions do you want the simulation to answer?•Sample questions:–How much will replacing our Ethernet hub with an FDDI hub increase throughput?–How much does a customized protocol improve efficiency in my cellular network?–How much will adding 1000 users lengthen response time?8Beijing Credit Top Co. ,Ltd. All rights reservedChoosing Aspects to Be Modeled•From the general questions, formulate specific research objectives, such as–Quantifying the throughput at a specific receiver.–Measuring the effect of changing a link’s transmission rate.–Finding the point where traffic load causes the system to become unstable.•These objectives determine the “granularity”(level of detail) required in the model.9Beijing Credit Top Co. ,Ltd. All rights reservedChoosing the Granularity Required•Choose the coarsest granularity that still answers all questions accurately.•Often, some portions of a model require fine granularity, and others much less.•OPNET is flexible --it allows you to vary the granularity within a single model.10Beijing Credit Top Co. ,Ltd. All rights reservedDefining the Data Input and Output•Input may refer to a predefined aspect of the model (such as connectivity) or may be more parametric (such as traffic generation rate).•To study effects on a system, keep most inputs constant and vary one or two over a range.•Determine which output (throughput, ETE delay, etc.) you need and the best presentation method (graphs, tables, animations, etc.)11Beijing Credit Top Co. ,Ltd. All rights reservedOutput may be presented as:graphs,animations,tables, web reportsor numerical values.Defining the Data Input and Output12Beijing Credit Top Co. ,Ltd. All rights reservedSpecifying the System Model•Simulation software can represent your system in many different ways. To use it effectively–Understand the features your simulation software offers.–Understand how those features can best describe your system.13Beijing Credit Top Co. ,Ltd. All rights reservedChoosing a Range of Input•Keep most variables constant except for one or two that you vary over a range.•Usually it is possible to analytically estimate upper and lower bounds for the range.•Choose input values that span this range.•After running initial simulations, you may decide to –Extend the range.–Focus on a narrow range to obtain finer detail.14Beijing Credit Top Co. ,Ltd. All rights reservedChoosing a Range of Input —Example•First range chosen was 2.0 -2.9 Mb/s with step 0.1 Mb/s.•Second range chosen was 2.4 -2.6 Mb/s with step 0.02 Mb/s.First range chosenSecond range chosen15Beijing Credit Top Co. ,Ltd. All rights reservedSystem Results Accurate?•Output should be “sanity-checked”for accuracy and robustness:–Does all output make sense?–Can it be justified?–Does the model behave appropriately?•One technique: predict output first, then compare predicted to actual output.16Beijing Credit Top Co. ,Ltd. All rights reservedResults Sufficiently Detailed?•As needed, either widen the input range to get “the bigger picture”or narrow it to focus on a specific region.•Should you–Collect more or different statistics?–Change the level of granularity?–Model different aspects of your system?17Beijing Credit Top Co. ,Ltd. All rights reservedResults Statistically Useful?•Is the model operating in steady state?–Expect initial transient behavior, converging to steady state.–Look for rising averages, increasing queue lengths, or other output that is still varying. This might mean that the model is not in steady state.18Beijing Credit Top Co. ,Ltd. All rights reserved Results Statistically Useful?•Run simulations for a longerperiod until outputs arestable.•A model that never reachessteady state indicates anunstable system.•Look at various statisticalmeasures to ensure that themodel is in steady state.19Beijing Credit Top Co. ,Ltd. All rights reservedResults Statistically Useful?•Have you run enough simulations?–Run numerous simulations, varying the random number seed and generating confidence intervals.–Run simulations until confidence intervals are tight.Beijing Credit Top Company Co., Ltd.北京怡嘉行科技有限公司Events and EventList Concepts21Beijing Credit Top Co. ,Ltd. All rights reserved Agenda•Conceptual Goals–Events–Interrupts–Event-driven simulation–Event list and the simulation time clock–Simulation Kernel–Processes and interrupts22Beijing Credit Top Co. ,Ltd. All rights reservedEvent-Driven Simulation•An event is a request for a particular activity to occur at a certain time.•OPNET simulations are event-driven. Time, in the simulation, advances when an event occurs.•A different method might be to sample at regular intervals. Disadvantages are as follows:–Accuracy of results is limited by the sampling resolution.–Simulation is inefficient if nothing happens for long periods.23Beijing Credit Top Co. ,Ltd. All rights reservedTime Event Type Module0.0 Initialize src.gen0.0Initialize src.rte4.3Timer expires src.gen4.3Packet arrives src.rteHeadEvent List Concepts•An OPNET simulationmaintains a single globalevent list.•All objects access a sharedsimulation time clock.•Events are scheduled onthe list in time order. Thefirst event on the list is thehead.•An event has dataassociated with it.•When an event completes itis removed from the list.24Beijing Credit Top Co. ,Ltd. All rights reservedInterrupts•An event becomes an interrupt when it reaches the head of the event list and is delivered by the Simulation Kernel to the designated module.•Data associated with the event can be obtained by the module when the interrupt occurs.•Certain modules, processes, and queues can be selected to place initial interrupts on the event list.25Beijing Credit Top Co. ,Ltd. All rights reservedThe Simulation Kernel•An entity, the Simulation Kernel (SK), manages the event list.•The SK delivers each event, in sequence, to the appropriate module.•The SK receives requests from modules and inserts new events on the event list.26Beijing Credit Top Co. ,Ltd. All rights reserved How Does the Event List Work?New event reaches head ofevent list, which causesSimulation Kernel to deliveran interrupt to the appropriatemoduleSimulation Kernel regainscontrol from moduleProcess, within themodule gains controland processesinterruptSimulation Kerneldeletes event fromevent list, allowingnew event to reachhead of list27Beijing Credit Top Co. ,Ltd. All rights reservedEvent List Implementation•The Simulation Kernel uses a proprietary, efficient algorithm to maintain the event list.•Event times are expressed as double-precision, floating point numbers and are used to keep the event list sorted.0.01234 56789 11111 110.01234 56789 11111 220.01234 56789 11111 330.01234 56789 11111 440.01234 56789 11111 550.01234 56789 11111 660.01234 56789 11111 77Suppose that this interrupt triggers an event to occurat 0.01234 56789 11111 75.The Simulation Kernel minimizes the time requiredto place this event at the correct place on the list.28Beijing Credit Top Co. ,Ltd. All rights reservedSimultaneous Events•What Happens When Two Events Are Scheduled for the Same Simulation Time?•The events are actually simulated sequentially, though they appear to occur synchronously according to the time clock.•The Simulation Kernel uses one of two methods to determine execution order:–“Natural order method”: The event that reaches the list first isexecuted first.–“Priority factor method”:Modules and events are assigned priority factors.Events with a higher priority or originating from a higher priority module are executed before events with a lower priority or from a lower prioritymodule.29Beijing Credit Top Co. ,Ltd. All rights reservedEvent List Concepts Reviewed•Some events must be entered on the event list at the start of a simulation.–A generator module enters an initial event.–A processor or queue module has the begsim interruptattribute enabled.•An event list typically has a few events -each event spawns another event or two that is placed on the list as the spawning event is deleted.•The event list is always growing and shrinking.•An event is pending until executed.A pending event can be cancelled.30Beijing Credit Top Co. ,Ltd. All rights reservedDelivery of Interrupts•When an interrupt is delivered to a module, control passes from the Simulation Kernel to the module.•If the module is a queue or processor, the interrupt is delivered to the process running within the module.31Beijing Credit Top Co. ,Ltd. All rights reservedForced States•Forced (green) and unforced(red) states differ significantlyin execution timing.•In a forced state, the process:–Invokes the enter executives–Invokes the exit executives–Evaluates all conditionstatements–If exactly one conditionstatement evaluates to true,the transition is traversed tothe next state.Transition to next stateTransition to next stateForced (green)statesEnter execsinvokedNo blocking or waitingExit execsinvokedEnter execsinvokedExit execsinvokedNo blocking or waiting32Beijing Credit Top Co. ,Ltd. All rights reservedUnforced States•In an unforced state, theprocess–Invokes the enter executives–Places a marker at the middleof the state–Releases control to theSimulation Kernel andbecomes idle–Resumes at the marker andprocesses the exit execswhen next invokedStart of invocationEnd of invocationUnforced (red) statesTransition to next stateBlocking, waitingfor invocationExit execs processedwhen invocationoccursEnter execsinvokedNextinvocationstarts hereBlocking, wait fornext invocation33Beijing Credit Top Co. ,Ltd. All rights reservedTransitions Between States•After completing the exit executives, the process evaluates the condition statements of all departing transitions from the state.•One and only one condition statement must evaluate to true.•The process traverses the transition associated with this condition statement.•A transition with condition = “default”is true if and only if no other conditions are true.•A transition with no condition set is termed unconditional and is always true.34Beijing Credit Top Co. ,Ltd. All rights reservedHow a Process Handles an Interrupt •Flow diagram showing how a process handles an interrupt(except the initial interrupt)Implementexit execsSet marker;block and waitfor interruptReceiveinterrupt EvaluateconditionstatementsRed state?YesFollowtransition tonext stateImplemententer execsFindmarkerNo35Beijing Credit Top Co. ,Ltd. All rights reservedProcess Model Example•Model with three forced states and one unforced state3. Transition occurs. 6. Transitionoccurs.2. Exit execs invoked immediately.Transition condition(pk_count==0)evaluates to true. 5. Exit execs invoked immediately.8. Marker is placedand process stopshere.7. Enter execs invoked.4. Enter execs invoked.1. Initial interrupt delivered and theenter execs invoked.36Beijing Credit Top Co. ,Ltd. All rights reserved Event List Example •Consider the following modelNetwork modelNode modelNode modelNode model: srcNode model: dest1Network model37Beijing Credit Top Co. ,Ltd. All rights reservedEvent List Example•The network model has three nodes (src, dest1,dest2) relying on two node models (both dest nodes use the same node model).•In the src node model, packets are generated at gen and sent by queue to either transmitter (tx0 / tx1).•Packets then flow across a link to a destination node (dest1, dest2) where they are received (rx) and thrown out (sink).•Three modules (gen, queue, and sink), have process models associated with them.38Beijing Credit Top Co. ,Ltd. All rights reservedThe BEGSIM Interrupt•BEGSIM is a special type of interrupt that occurs at simulation time 0.0, before any other type of interrupt.•A BEGSIM interrupt usually initializes a module and schedules future events.•Any processor or queue can have its begsim intrpt attribute enabled, resulting in an event being placed on the event list for time 0.0.39Beijing Credit Top Co. ,Ltd. All rights reservedEvent List Example•The begsim intrpt attribute for gen and sink is enabled; this places two events on the event list.Time Event Type Module0.0BEGSIM src.gen0.0BEGSIM dest1.sink0.0 BEGSIM dest2.sinkNode model40Beijing Credit Top Co. ,Ltd. All rights reservedProcessing the First Interrupt •Consider the process model specified by the gen module.Node model: src Process model: genStarting the Simulation1. To start the simulation, the head of the event list isprocessed and a BEGSIM interrupt is delivered to the process in src.gen.2. Because this is the first interrupt, the process beginsexecution at the initial state, marked with a black arrow. 3. Because this state is being entered, the enter execs are executed.Time0.0 0.0 0.0Event TypeBEGSIM BEGSIM BEGSIMModulesrc.gen dest1.sink dest2.sink41Beijing Credit Top Co. ,Ltd. All rights reservedProcessing the First Interrupt at genProcess model: genProcess model42Beijing Credit Top Co. ,Ltd. All rights reservedProcessing the First Interrupt at gen4. Process invokes and completes the enter execs of Init. 5. Because Init is a forced (green) state, process immediately invokes and completes the exit execs. 6. Process evaluates all condition statements. This state has only one departing transition which evaluates to true. 7. Process transitions to Wait.43Beijing Credit Top Co. ,Ltd. All rights reservedProcessing the First Interrupt at gen8. Process invokes and completes the enter execs of Wait.Process model44Beijing Credit Top Co. ,Ltd. All rights reservedProcessing the First Interrupt at gen9. One action in the enterexecs (line 8) is to schedule a self interrupt by means of a KP. This adds an event to the event list.Time0.0 0.0 0.0 4.3Event TypeBEGSIM BEGSIM BEGSIM SELFModulesrc.gen dest1.sink dest2.sink src.gen10.Process places amarker at the middle of Wait. 11.Process becomes idle.Marker45Beijing Credit Top Co. ,Ltd. All rights reservedProcessing the Second Interrupt at sink12.Simulation Kernel removes the first event and advancesthe next event to the head of the event list. The simulation time remains 0.0.13. Simulation Kernel delivers a BEGSIM interrupt to dest1.sink. 14.Process in sink module gains control. It executes until it reaches an unforced (red) state, places a marker, and then becomes idle. (This model is not shown.)TimeFirst event is removedEvent TypeBEGSIM BEGSIM SELFModuledest1.sink dest2.sink src.gen0.0 0.0 4.346Beijing Credit Top Co. ,Ltd. All rights reservedProcessing the Next Interrupt at gen15.The Simulation Kernel removes the previous event andadvances the next event to the head of the event list. The simulation time becomes 4.3 seconds.16.The event at the head of the list causes a SELFinterrupt to be delivered to the src.gen process. The process resumes at the marker in the middle of Wait.17.Process invokes and completes the exit execs of Wait.Time4.3Event TypeSELFModulesrc.gen47Beijing Credit Top Co. ,Ltd. All rights reservedContinuing the Process at gen18.To leave Wait, the process evaluates all condition statements. This state has one outgoing transition and its condition statement (shown) evaluates to true.48Beijing Credit Top Co. ,Ltd. All rights reservedContinuing the Process at gen19. The process transitions to Send.20. The process invokes the enter execs of Send and calls op_pk_send ( ) to send a packet. This results in an event of type STRM being placed on the event list.Time4.3 4.3Event TypeSELF STRMModulesrc.gen src.queue49Beijing Credit Top Co. ,Ltd. All rights reservedContinuing the Process at gen21. Because Send is a forced (green) state, the processimmediately invokes and completes the exit execs.22. Process evaluates all possible transitions. One evaluatesto true.23. Process transitions to Wait. 24. Process invokes enter execs of Wait , schedules anotherSELF interrupt in the event list and becomes idle. 25. Simulation Kernel takes control and processes the next event on the list.50Beijing Credit Top Co. ,Ltd. All rights reservedSimulation Termination• Simulations terminate in one of four ways–The event list is emptied. –Simulation attribute duration expires. –A process calls for termination, using the KP op_sim_end(). –A fatal error occurs.51Beijing Credit Top Co. ,Ltd. All rights reservedHow Does Time Advance?• Simulation time advances only when an event with a latertime is taken from the event list.• No simulation time occurs during an invocation of a processmodel.• No time elapses during transitions between states.–A process model must always end in a red state so time can advance. –Avoid endless looping between forced (green) states.52Beijing Credit Top Co. ,Ltd. All rights reservedSummary• Forced and unforced states differ in when they execute theirinstructions.• Any processor or queue can have the attribute begsimintrpt enabled, scheduling an event for time 0.0.• Control passes dynamically between the Simulation Kerneland individual processes.53Beijing Credit Top Co. ,Ltd. All rights reservedBreak54Beijing Credit Top Co. ,Ltd. All rights reservedBeijing Credit Top Company Co., Ltd.Process Modeling Methodology (PMM)北京怡嘉行科技有限公司Introduction• What is the Process Modeling Methodology (PMM)?– A systematic approach to creating process models in OPNET Modeler – Quickest, most efficient method of development – Protection from some common pitfalls – Produces consistent results – Indispensable for larger models• But first…What is a Process Model?56Beijing Credit Top Co. ,Ltd. All rights reservedOPNET EnvironmentProcess model is the heart of custom simulation development57 Beijing Credit Top Co. ,Ltd. All rights reservedComparing Approachesgg gggg ggFlower Petal - tcp_manager_v3.pr.mPMM - tcp_conn_v3.pr.mBoth of these are parts of the standard OPNET TCP modelBeijing Credit Top Co. ,Ltd. All rights reserved58Process Model DynamicsinitINITIALIZED MSG1msg1 gMSG2connectCONNECTEDREGISTEREDregisterFAIL FAILwaitMSG3msg2 gFAILRECOVERfailBeginners avoid RED states, but they are very important!59msg3 gBeijing Credit Top Co. ,Ltd. All rights reservedStates, Events, and ActionsImportant definitionsStateEventActionStateAlarmasleep/WakeupawakeCondition of system0:0060Stimulus to move system6:00InstantaneousResponse of system to stimulus6:00Condition of systemTime progressesTime progresses8:00Beijing Credit Top Co. ,Ltd. All rights reservedSteps for Building a Process Model• Obtain protocol specification • Design using the methodology • Review design • Implement in OPNET61Beijing Credit Top Co. ,Ltd. All rights reservedProcess Modeling Methodology Stages• First four stages: Design– Stage 1: Context definition – Stage 2: Process level decomposition – Stage 3: Enumeration of events (per process) – Stage 4: Event Response Table development (per process)This is an iterative stage and is the meat of process model methodology4 1 2 3• Last stage: Implementation in OPNET– Stage 5: Specification of process actions (per process)562Beijing Credit Top Co. ,Ltd. All rights reservedProcess Modeling Methodology Example 1: Stop-and-Wait Retransmission Protocol• Protocol for the Data Link Layer at the sending side only • Basic functionality: Provide reliable communications over alossy channel• Accept frames from the upper layer and send them to thephysical layer63Beijing Credit Top Co. ,Ltd. All rights reservedProcess Modeling Methodology Example 1: Protocol RequirementsAmalgamated Communications Kompany (ACK) C. R. Ash SWaRe Protocol Version 1.3.X Aug 23, 2005 ProprietaryReceive frames from higher layer Transmit frame over physical layer Must receive an acknowledgment before sending the next frame Queue frames arriving from the higher layer if the process is waiting for acknowledgment for a previous frame Retransmit the frame if an acknowledgment is not received before the end of a timeout period When the link fails, frames cannot be sent or retransmitted until the link recoversBeijing Credit Top Co. ,Ltd. All rights reserved64Just for Fun…How many lines of OPNET code are needed to model this protocol?65Beijing Credit Top Co. ,Ltd. All rights reservedStage 1. Context Definition• Fit this piece into the big picture: • Step 1: Identify interdependent modules • Step 2: Select communication mechanisms withinterdependent modules4 1 2 3• Step 3: Develop diagram of system andinterdependent modules566Beijing Credit Top Co. ,Ltd. All rights reservedStage 1: Context Definition (cont.)1 2 3Higher LayerFrames for Transmission4SWaRe ProtocolTransmitted Frames AcknowledgementsPhysical Layer567Beijing Credit Top Co. ,Ltd. All rights reservedStage 2: Process-Level Decomposition• Step 1: Determine which process modeldecomposition technique is applicable to the system– Single process model – Multiple (parent-child) process models1 2 3• Step 2: If a multi-process implementation is chosen,identify the areas of responsibility assigned to each process4• Step 3: For multi-process implementations, determinecircumstances of process creation and which process will be the root568Beijing Credit Top Co. ,Ltd. All rights reservedStage 3: Enumeration of Events• Step 1: Define logical events of each process • Step 2: Select event implementation methods1 2 34 Frames for TransmissionAck timeoutSWaRe Protocol! !System - Link Fail System - Link Up 5Acknowledgements69Beijing Credit Top Co. ,Ltd. All rights reservedStage 3: Enumeration of Events (cont.)Event Name Event Description Interrupt Type1 2 34570Beijing Credit Top Co. ,Ltd. All rights reservedStage 3: Enumeration of Events (cont.)Event Name Frame Arrival Event Description Interrupt Type A frame has arrived Stream from the higher level. The retransmission timer has expired. Self4 1 2 3Timeout Acknowledgement Received! !An Stream acknowledgement for the last transmitted frame has been received. The link has failed. The link has recovered. The process starts. Failure Recovery Begin Simulation5Link Fail Link Up Power Up71Beijing Credit Top Co. ,Ltd. All rights reservedStage 4: Event Response Table Development• The Event Response Table specifies how the model reactsto various events, depending upon the state of the modelState Event Condition Action Final State1 2 34• The events and actions are indicated in the specification • What are the conditions? • What are the states?72 Beijing Credit Top Co. ,Ltd. All rights reserved5ConditionsState Event Condition Action State73Beijing Credit Top Co. ,Ltd. All rights reservedStage 4: Event Response Table Development - Identifying States• A state…– Represents a blocking point of the process – Corresponds to particular sequences of events having occurred – Reacts to events in a specific manner – Is mutually exclusive of and complementary to other states4 1 2 3• Problem:– Typically we do not know all possible states of the system at this point in the design process5• Solution:– Start with an initial state and “walk” through the model74 Beijing Credit Top Co. ,Ltd. All rights reservedStage 4: Event Response Table Development: Iteration Plan - Start With Initial State• Step 1: Choose a state–Step 2: Choose an eventStep 3: Choose a condition under which the event occurso Step 4: Determine all actions to perform o Step 5: Determine the final state1 2 3Loop Step 3 for all conditions–Loop Step 2 for all events4• Loop Step 1 until all states are completeState InitBeijing Credit Top Co. ,Ltd. All rights reservedEventConditio Action nFinal State575Stage 4: Event Response Table Development - Initial StateUsing a BEGINSIM interrupt assures that the Init state will not have to handle any other type of event 1 2 3State InitEvent Frame ArrivalConditio nActio nFinal State4Timeout Acknowledgeme nt Received Link Fail Link Up Power Up765AlwaysNoneIdleBeijing Credit Top Co. ,Ltd. All rights reserved。