Class Note for EECS 563 with Professor Frost at KU (2)
Class Note for EECS 563 with Professor Frost at KU (2)
Popular in Course
Popular in Department
This 26 page Class Notes was uploaded by an elite notetaker on Friday February 6, 2015. The Class Notes belongs to a course at Kansas taught by a professor in Fall. Since its upload, it has received 16 views.
Reviews for Class Note for EECS 563 with Professor Frost at KU (2)
Report this Material
What is Karma?
Karma is the currency of StudySoup.
You can buy or earn more Karma at anytime and redeem it for class notes, study guides, flashcards, and more!
Date Created: 02/06/15
Communications Network Simulation 8 Victor S Frost Dan F Servey Distinguished Professor Electrical Engineering and Computer Science University of Kansas 23351rving Hill Dr Lawrence Kansas 66045 Phone 785 8644833 FAX785 8647789 email frosteecs kuedu httpWwwittc kuedu 1 With modi cation from Dr David Petr Simulation Outline Define network simulation Discuss attributes and application of simulation Present implementation of simulation systems Discuss analysis of simulation results Discuss selection of simulation tools Provide an overview of Extend the simulation tool using here 2 Simulation A Definition of Communication Network Simulation Communication network simulation involves generating yseudorundom seguences of message lengths and interarrival times or other input processes eg time varying link quality then using these sequences to exercise an algorithmic description of the network oyerution 3 Simulation Attributes of Simulation I Simulation Is a Very Flexible Evaluation Tool gt General Network Characteristics Sources Topology Protocols Etc gt Minute Detail Simulation Models Can Be Expensive to Construct gt Human Effort Simulation Models Can Be Expensive to Run gt Computer Effort Statistical Analysis of the Results Can Be Difficult gt Requires Careful Interpretation Difficult to Gain Insight Into System Behavior gt Simulate Only a Set of Speci c Scenarios 4 Simulation When to Use Simulation I Whenever Mathematical Analysis Is Difficult or Impossible gt For Studying Transient Behavior of Networks gt For Systems With Adaptive Routing gt For Systems With Adaptive Flow Control gt For Systems With Blocking Finite Buffers gt For Systems With General Message Interarrival Statistics Simulation When to Use Simulation I For Validating Analytic Models and Approximations gt How Accurate Is the Model gt Do Approximations Distort the Results I For Experimentation Without Disturbing an Operational System gt Test Possible Modifications and Adjustments Simulation Modeling Elements for Communication Networks I Traffic and Input Processes gt Message Arrival Process 7 Often Interarrival Times gt Message Lengths gt Other Message Attributes 7 Service Class 7 Error models I Algorithmic Descriptions of Network Processing gt Protocols gt Links and Queues gt Routing Simulation 7 Sample Realization of an Input Process Table Message number 1 2 3 4 5 6 7 8 9 10 11 12 lnterarrival time 2 1 3 1 l 4 2 5 1 4 2 between i1 and i message seconds Length of it1 message 1 3 6 2 1 1 4 2 5 1 1 3 seconds Simulation Sample Realization of an Input Process Graph i275 Measured Video Traffic 30 Arrival 95 Events amp 7 129 Lengths g was was 395 12m 12 Time 9 Simulatt n Discrete Event Simulation Terminology I Entities gt Objects Upon Which Action Is Performed gt In Network Simulation Entities Are Messages Packets I Attributes gt Characteristics Which Describe Entities eg Message Length I Events gt Occurrences That Trigger Activities eg Message Arrival Departure 10 Simulatt n Discrete Event Simulation Terminology I Activities gt Operations That Change the State of the Network gt Example Increment Number of Messages Waiting in a Buffer I Files gt Groupings of Entities Which Share a Common Attiibute gt Example All Messages Waiting in Buffer 11 Smuhmn Discrete Event Simulation Dynamics Network 5mg m Amoin WWW sum Acovity Network 5mg 12 Smuhmn Time Step Approach to Network Simulation I Approaches to Discrete Event Simulation gt Time Step Approach Fixed Increment Time Advance gt EventScheduling Approach I Fixed Increment Time Advance gt Choice of Increment Important gt Too Large Multiple Events Happen In Single Step gt Too Small Wasted Processing Time gt Update System States at End of Each Fixed Time Interval Simulahan Time Step History of Simple Statistical Multiplexer mwlmmwa sum mm mm Numb m W Ami m m m m 2m 5qu mm gym 0 Q l o i D D a 9 1 0 2 Q Q a l cg 2 lt1 1 2 4 G lt2 2 2 s o i 2 2 a co 2 3 7 g 2 3 3 3 4 9 s 4 m 3 4 u g Q 2 z 5 50 l2 Q 3 4 a is 2 3 a 7 l4 c9 g 2 a a 7 is o 1 2 3 7 is 2 w l 2 is 2 is a 5 lt3 i 2 523 via 20 2 3 ii 3 m y z 213 703 22 l 2 21 l 2 24 o 2 3 25 z a 14 26 0 a 2 z s we 5mm Number in System Nz vs Time Number in system Plotter Discrete Event 4533333 4 166657 3775 3333333 2916657 25 l 3 2 083333 Mi 1 666667 125 Simulation Event Scheduling Approach to Network Simulation I Variable Time Advance gt Advance Time To Next Occurring Event I Update System State Only When Events Occur gt For Example Arrivals or Departures 16 Simulation Event Scheduling Approach to Network Simulation I Event Calendar gt Events Instantaneous Occurrences That Change the State of the System gt An Event is Described by The Time the Event is to Occur The Activity to Take Place at the Event Tilne gt The Event Calendar is a Time Ordered List of Events 17 Simulation Event Scheduling Approach Simplified Flow Control gt Use Event List An Executive to determine next 01 Majnljne event to process Controls the Selection Of Next Advance simulation Event clock to event time 1 Update system state using event routines Update event list using event routines 18 Simulation Event Scheduling for Simple Statistical Multiplexer Arrival 19 Event Scheduling for Simple Statistical M u lti plexer Deparwre End of Transmissio gt vs aquot NO emDLl Change Read 8 remav status of from file transmissior the attributes faculuty of next to send Return 20 Snapshot of Simple Statistical Multiplexer Simulation Stalus of Tiansmlssmn I acme Message numbe Mesane lengm Message Inle 4 2 E Numbev m Guile annex Cumums Achwty lype Arrival I Departure H llnle Fvcnl Cnlcllum um ljel ul Amlividm vigil 21 Simulation 9 Approach Advantages Disadvantages Time Step Efficient for system with Very Must process at each time step frequently occurring events Error induced by fixed finite Efficient for regularly spaced time increment events Must establish rules to order events that occur in same time increment Event Scheduling Only process at event times Significant programming effort required No time increment to select Flexible 22 Simulation Verification and Validation of Simulation Models I Model gt Mathematical Algorithmic Description of Behaviour of Real Thing I Verification gt Determining Whether the Simulation Model Performs As Intended gt In Programing Terminology Debugging gt Example Is VIW1 Model Producing Exponential Message Lengths I Validation gt Determining Whether the Simulation Model Itself Is an Accurate Representation of the Communication Network Under Study the Real Thing gt Example Is the Assumption of Exponential Message Lengths Accurate 23 Simulation Verification Methods I Modular Development and Verification gt Break Large System Into Smaller Components gt Verify Component by Component I Structured WalkThrough gt Step by Step Analysis of Behavior for Simple Case 24 Simulation Verification Methods I Event Trace gt Detailed Analysis of Model Behavior gt Compare to Walk Through Analysis I Model Simplification and Comparison to Analytic Results I Graphical Display of Network Status As the Model Progresses gt To See What Is Happening As It Happens 25 Simulation Some Comments on Validation I Silnulation Models Are Always Approximations I A Simulation Model Developed for One Application May Not Be Valid for Others I Model Development and Validation Should Be Done Silnultaneously I Specific Modeling Assumptions Should Be Tested I Sensitivity Analysis Should Be Performed I Attempt to Establish That the Model Results Resemble the Expected Performance of the Actual System I Generally Validation Is More Difficult Than Verification 26 Simulation Analysis of Results Statistical Considerations I Starting Rules gt Overcoming Initial Transients gt An Initial Transient Period Is Present Which Can Bias the Results gt Achieving Steady State Use a Runin Period I Determine Tb Such That the LongRun Distribution Adequately Describes the System fort gt Tb Use a Typical Starting Condition State to Initialize the Model I Quality of Performance Estimates gt Variance of Estimated Performance Measures 27 Simulation Quality of Performance Estimates I Performance Estimates Should Be Unbiased I Performance Estimates Should Have Acceptable Variance I Confidence Intervals Are the Usual Approach for Assessing the Accuracy of the Estimators I The Desired Confidence Interval Width Determines the Length of the Simulation Run I Observations Tend to Be Correlated gt Cannot Directly Apply Standard Statistical Approaches Based on iid Independent ldentically Distributed Observations 28 Simulation Dealing with Lack of Independence I Simple Replication Multiple Simulation Runs gt Assume Results for Each Replication Are Independent gt Inefficient Because of Multiple Startup Periods I Batching Divide Single Simulation Run Into EqualTime Batches gt Assume Results From Each Batch Are Independent gt Batches Can Be Correlated Unless Dead Periods Between Batches Are Employed 29 Simulation Simple Replication Approach O Nw u l I Tb A No Fl rl LIJ LLLH lUH IJL Replication l A Nu i Ii i i f ij 2 Tb i Nlttgt 39 l rLlJ Ll il39l l ll IJ m Ll M Tb 30 Simulation O Imwgm odmw um I Batching Approach mt Nt S 4 3 2 1 o gtt d Batch 1 D gt lquot39 Balch M l T T 2T MT b 1 1 1 31 simulation Evolution of Network Simulation Tools quotZerothquot Generation 7 General Purpose Languages gt Fortran C c Pascal Basic quotFirstquot Generation 7 General Purpose Queueing Systern Sirnulations gt GPSS SLAM SIMSCRIPT quotSecondquot Generation 7 Application Specific Cornputer Systems and WiderArea Communication Networks gt RESQ PAWS Nsez http www 15 edunsnarnns quotThirdquot Generation 7 Integration of Second Generation Languages With a GraphicscOriented Analysis and Modeling Environment gt Extend WWWinaginethatinCCom gt SEEWorkbench Scienti c and Englneenng Software Inc Austin TX gt OI N39ETI1113 Inc Washington DC 32 simulation Relative Merits of General Purpose Languages Advantages Disadvantages Wide Availability Longer programming and debugging time Few restrictions imposed on the model Difficult verification User may have prior knowledge of the language Unless objectoriented limited ability to reuse models Generally more computationally efficient Model enhancement and evolution are difficult 33 Slmulahon Relative Merits Purpose Langu Cdvanuges Provide huilteln siniuldtion services to rvducu prugrarnming offuri ofSpedal ages 39 Tavmuges Niusl adhere to a particular world viewquot ut the language Provide errorrcliccking techniques superior to thine provided in general purpose languages Availability and support Provide a brief direct vehicle tor L39xpr sing the concepts arising in a Sli39nulntiun study Cost Provide aliilil to Cum rct ueer subroutines requier a part of any Sll nlll tlun routine Contain set if subruulines hir Column randon39i nui nbvrs Increased computer running tilne 8 required to learn the ge and niodvlulg paradigm Facilitate collectinn and display or data produced Facilitate mod cl reuse Slmulahon Relative Merits of ComputerAided Analysis and Design Environments Advantages Disadvantages Provide a complete integrated performance analysis environment Tailored to a specific modeling paradigm Graphically based May be tied to a specific hardware platform Typically integrate language database prior knowledge and statistical analysis packages Increased execution time Support management of models and input output data Cost Facilitate model reuse and group model development Simulation 35 Criteria for Selecting a Network Simulation Tool Availability Cost Usage Documentation Ease of Learning Computation Efficiency Flexibility Portability User Interface Extendibility Memory Requirements 36 Simulation Guidelines to Network Modeling and Simulation I Things to Know gt Know the Customer gt Know the Network gt Know the Important Performance Metrics I Things to Do gt Establish a Credible Model gt Expect the Model to Evolve gt Apply Good Software Management Techniques Simulation 37 Conclusions I Simulation Can Be an Important Tool for Communication Network Design and Analysis I Care and Thought Must Go Into Construction of Communication Network Models I Care and Thought Must Go Into Interpretation of Model Output Simulation 38 Extend Overview I Allows Graphical Description of Networks gt Sources Links Nodes Etc I Data Flow Block Diagrams I Hierarchical Structure to Control Complexity I Be sure and create libraries when creating complex models 39 Simulation Extend Simple Statistical Multiplexer Model Simulation Clock Executive FIFO Queue Exit Required 40 Simulation Extend Better Statistical Multiplexer Model Packetsuume unx Capacity 1 ms mun Arrival Rate Packetssec Message Length bits Meaasge Length in bll Message Length Simulation 41 Extend Better Statistical Multiplexer Model Meaasge Length lrl bits Gettirne in System Simulation 42 Extend Data Structures mum m human Maw wan rm m rvmunu 31 manna mum Extend Block Diagram H iera rchy m m Suurce mum Extend Packet Switching System Packet Switching System Packet Source 46 Simulation Packet Switching System Packet Switching System Output Port PacketOut Link Capacity bs Simulation Simulation Case Study Simulation ofATM WAN s I Determine the level of model fidelity required to accurately predict ATM WAN performance I Determine the feasibility of measurement based validation of TCP IP over ATM WAN simulation models I Identify factors in uencing TCPIP over ATM WAN performance 49 Simulation Simulation Case Study Simulation of A W WAN s System Parameter Value TCP MTU size 9180 bytes TCP processinq and OS overhead time DEC 3000 AXP 200300 us SGI 550 us SPARC 10 550 us SPARC 5 700us TCP user send buffer size 64 kates Slowtimer period 05 s Fasttimer period 02 5 Minimum RTO 105 AAL5 SAR processing time 02 us AAL5 cell payload size 48 Bytes SNitch processinq time 4 us SNitch output buffer size per VC 256 cells OC3c link speed 155 Mbs TAXI link speed 100 Mbs DS 3 link speed 45 Mbs 50 Simulation Simulation Case Study Simulation ofATM WAN 5 Network Con guration mm M aquot mm m 54mm mm 7 mm m um 1 km W A hrmm mm m m mm nt 3 m2 ISS IM nu minim um Hth mint mm m M HALmt SPRINT ATM Public Nmmrk A u xuw s 51 hm Comparison of Experimental and Simulation Performance Predictions
Are you sure you want to buy this material for
You're already Subscribed!
Looks like you've already subscribed to StudySoup, you won't need to purchase another subscription to get this material. To access this material simply click 'View Full Document'