Network Design and Modeling - Lab II: Travaux Pratiques : Simulations ns2
M1 – Sorbonne Université Network Design and Modeling
Lab II: Implementation of a simulation model and its application – follow-up
In this second part of the Lab, you will study some performance metrics of the previous implemented network with multi- routes by running simulations and also by modeling.
IMPOR TANT . At the end of this lab (4:00pm), you must send a first version of your lab report. This first version will include the simulation results and some texts that explain them . Afterwards, you will write and send a final report. This final report must complete the texts and include additionally a comprehensive analysis for each result in which you explain carefully your simulation observations in order to provide useful conclusions. Besides, answers to questions 4) to 9) must be complete and well developed. Any copying-pasting or plagiarism will be sever ely punished.
The dead line to send this final report is Tuesday , April 25, 2023 befor e 4:00 pm. Of course, you can send the final report befor e or even at the end of this session.
Recall of the network scenario with multi-routes and losses
The network that you implemented in the first part of this lab is composed by 4 nodes , and 4 links . Packets of 1000 bytes are generated by a Poisson source at node 0. The sending rate is b in kbits . The destination of these packets is node 3 (see the figure below). ¼ of the traffic is routed through node 2. Links 1-2 and 2-3 may drop packets with the probability p (Uniform losses ). All queues are FIFO . In addition, we assume that the application at node 3 cannot support more than 64 packets per second. We would like to study the maximum performance of the network as a function of b and p.
1) - BIS:
First provide in your report a screenshot of the network animation with nam showing clearly the two flows traveling through the two different routes. The screenshot must show some part of your specific desktop environment (so we can verify that it is yours)
Second, include the code of your TCL NS2 script in the report.
Before you move to the next questions, make sure to set all propagation delays to 0ms
2) Set p=0.1. Find by simulations the maximum rate bMAX sent from the source before the network becomes unstable (congested). To do so, run several simulations while varying the input rate b from small values to higher values by a step of 10 kbits . After each simulation, analyze the trace file with your own code that you will develop specifically for the purpose of computing the average delay of packets travelling from node 0 to node 3 of the figure above during the simulation. If you want, you can use the previous awk script analyze1.awk and modify it accordingly , or inspire from the awk script shown in Lecture 1 slides 13 and 14. Finally, plot the average delay as a function of b as shown in the figure below .
Remark 1: The simulation time must be long enough, at least 10000s. The seed of random number generation must be your student id as in the previous lab.
Remark 1: The regio n of bMAX must include more simulation points. For instance, if you find that bMAX is between 100 kbits and 110 kbits , then you must add points related to simulations with b=101, 102, 103, 104, 105, 106, 107, 108 and 109. (step of 1 kbits ). By doing this, you obtain a smoother curve and a better precision for the value of bMAX. Finally , plot with lines only (without points) for a nicer curve or if you want to keep points then reduce their sizes.
3) Now , set b = 100 kbits . Plot the packet loss ratio as a function of p. Again you should run several simulations while varying p. You can use a step of 0.05.
Questions:
- Determine p0, the dropping ratio that causes 5% loss ratio.
- Is it coherent that the loss ratio of 5% is lower than the dropping ratio p0? Why?
4) Now comp ute bMAX of the question 2) analytically . Hint: Locate the bottlene ck of the network. Justify carefully your answer .
5) Assume that b < bMAX. Find the analytical expression of the packet loss ratio tPP as a function of p. tP P = [(T otal input rate) – (T otal output rate)] / (T otal input rate) .
Then, using the found expression, compute the value of p0.
6) Compare the analytical results of questions 4) and 5) with those obtained via simulations in question 2) and 3).
7) Among the queueing models that you learned in the first part of NDM lectures, choose the better one, in your opinion, to model the studied network. Justify and explain your model while indicating its parameters. Plot on the same figure the delay as a function of the input rate b using your model, and the previous delay obtained in 2). Explain and analyze the dif ferences and the similarities between the two curves.
8) Repeat the simulations of 2) but this time you change all link queues limit to 5 (queue-limit ). Then, answer 7). In other words, try to find the better queueing model, in your opinion, to model the new studied network with queue limits. Justify and explain your model while indicating its parameters. Plot on the same figure the delay as a function of the input rate b using your new model, and the previous delay obtained in this question 8). Explain and analyze the dif ferences and the similarities between the two curves.
9) To finalize your report, discuss and compare generally the two approaches of performance evaluation: analytical modeling vs. simulations. When should we use modeling and when should we use simulations? When should we use both? You can use the previous Lab to complement your synthesis. Also, you can inspire from this book but do not copy-paste full sentences from it.
Note. For this lab, source codes of the trace analyzer and the plotter must be included in your pdf report, the first and the final one.
END.
