Network Simulation with Opnet

Simulation of Networks Computer experiment, alternative to building a network Input network and traffic details Specify performance stats to be collected Run the simulation Examine the results – use graphs, spreadsheet Evaluate averages, min/max, deviation

Input details Components – PCs, workstations, LANs, links, link speeds, protocols User profiles – applications, start/stop time, destination servers Application details – transaction interarrival time, transaction size ►Complex behaviour of networks Difficult to predict (even simple sub-networks and single links can have complex behaviour due to the large number of interacting dynamic aspects).

C/S Database Application – link load example Ratio of query to update 50-50 Mean (average) transaction interarrival time = Exponential (2 sec) (Constant, Exponential are statistical distributions – discussed later) Try calculating the load in bits/sec each way. Query (C to S): 128 bytes, Response (S to C) 2400 bytes Client (C) Server (S) Client to server: one query (128) and one update (2400) every 4 seconds. Total traffic = 8 * (128 + 2400) bits in 4 seconds = 5056 bits per sec. Update request (C to S) 2400 bytes, Confirmation (S to C) 128 bytes

Video conferencing bandwidth requirements example 30 frames per second Frame size: 800 x 640 pixels Pixel resolution 12 bits Load = [30 x 800 x 640 x 12] bits/sec = 184,320,000 bits / second (approx 200Mb/s)

Performance statistics - examples Output from the simulation experiment: Application response time, delay Voice/video packet end-to-end delay Network access delay No. of collisions Link/server utilisation, throughput Packet loss, retransmission

OPNET – Simulation Package Facilitates simulation of communications networks Enables determination of performance figures Is a complex and powerful package with many features: Get started by doing some of the tutorials. Then modify the library models, and compare different scenarios. Save your own version of the library model under a different name (your-initials_LANs). Save to your O: drive. Then develop your own models, code etc.

OPNET Start up screen Project may have multiple (related) scenarios. A scenario is a specific configuration of a model. You can open projects, nodes etc. from the file menu. Documentation and tutorials from the ‘help’ menu.

See ‘help’ menu for tutorials

Small Internetworks Introduction This tutorial shows how you can do organizational scaling to solve a typical "what if" problem. You will learn how to use OPNET Modeler features to build and analyze network models. The tutorial focuses on the use of the Project Editor, and how it will be used with the Node and Process editors in later lessons. Tutorial activities include: Build a network quickly Collect statistics about network performance Analyze these statistics

Tutorials for this course Start with: Small Internetworks: Create, simulate & analyse the (star network, client server) model within the project editor. Compare two scenarios. M/M/1 Queue: Build a simple queuing system and simulate it.

Workspace & documentation Opnet Workspace (various GUI components, inc.) Menu bar Action buttons Message display Tool Tips Documentation On line documentation (see ‘help’ menu) Comments within model components From the Opnet web site OPNET Website: http://www.opnet.com/rpm_support/

Work Sequence when using OPNET Create the network model – nodes, links, processes – always save to your drive (O:/) Specify traffic types (e.g. database, web browsing) and volumes (how many packets, what size packets) Specify stats to collect – e.g. delays, buffer use, queue size, load, utilisation for servers, links. Run the simulation – how long? Seed? Estimate performance parameters (utilisation, delays, throughput, buffer use, loss/error rates). Average, min/max, standard deviation Export to Excel spreadsheet for better analysis and graphs

Main Components of OPNET Project Editor: build a network model using various model libraries; specify statistics (results) to collect, run the simulation and examine results Node Editor: Nodes have modules (e.g. MAC, IP, TCP, ARP) & processes (module behaviour, described by State Transition Diagrams STD & Finite State Machines FSM). Process Model Editor: Process functionality (C++). Link Model Editor: Create data links, specify speed, protocol etc. Packet Format Editor: Create different types of packets & frames Probe Editor: To specify simulation statistics Simulation Tool: To run multiple simulations, with different run parameters, and other sophisticated features Analysis Tool: More detailed analysis, comparisons etc.

Scenarios The project describes the model Scenarios are slight variations of a basic model. E.g. Change in link speed, traffic parameters etc.

Object Attributes right click and edit attributes Attributes of node-30 (Central Switch) 3Com model, type etc. included in the name Explanation (‘details’) of attributes can be obtained

User Profiles Each profile represents a typical user – a mix of applications Start/stop times are defined for profiles

Applications Standard applications are defined Here is a list of what is available Actual parameters can be modified

Application Parameters

Example traffic specification Telnet (remote login) example shown on previous slide. Enter command (interarrival) time (secs) Command size (bytes) for terminal traffic Command size (bytes) for host traffic Distribution needed too – constant, exponential, uniform etc. Note / recall: Load (bits/sec) = commands/sec x command size (bytes) x 8 Commands/sec (rate) = 1/interarrival_time Always remember: If the interarrival time = t secs, Arrival rate = 1/t commands/packets/messages per second.

Running the simulation Run length – enough to generate enough packets. Values per statistic – I suggest 40-50 Seed – determines the sequence of random numbers used to generate packet arrival events etc. A given value of the seed will gave the same results every time. A different seed will give slightly, but not drastically different results Calculate the average over 3-4 runs with different seeds to get a more accurate estimate. Use optimised mode for quick execution.

Guidelines Duration or run length? Enough for 20k-30k messages of each type. Values per statistic: if duration = 3600 seconds, values/statistic = 60, each value will be averaged over 3600/60 = 60 seconds. Independent runs? Change the seed each time. Results from different runs will be close, but not identical. Enter parameters in the right units (e.g. seconds, bytes) Report results, using appropriate units (e.g. 5 ms rather than 0.005 sec).  

Results Graphs of delays, utilisation etc. can be obtained (Opnet graphs are poor quality for coursework reports) Graph properties can be changed if you wish You can also export to Excel spreadsheet to analyse, such as calculate averages, and also generate much better quality graphs

More Results Multiple graphs for comparison Separate (stacked) or overlaid (same frame) Various types of plots – as is, average, time average, histograms

Checking Results Estimate or calculate a sample of results to confirm your model, its parameters etc. are correct. Example: Message inter-arrival time = 10 seconds, message size = 1000 bytes. Selected output statistics: Messages/sec sent and/or received Bits/second sent and/or received Results should be: Messages/sec sent or received = 0.1 Bits/sec sent or received = 800

Example calculation of utilisation ( (Rho) ) Messages Pkt inter-arrival time 0.2s Pkt arrival rate = 1/0.2 = 5 packets per sec Pkt size = 1600 bytes Link/channel bitrate = 128 Kbits/s   Number of bytes per sec = 5 x 1600 Number of bits per second = 8 x 5 x 1600 Link (channel) utilisation = (8 x 5 x 1600)/128000 = 0.5 or 50%