Assignment 4 for csc8513, S2 2015
This assignment is worth 30% of the marks for the course. It requires you to perform various experiments in Netml.
In order to pass this assignment you will need to be observant. You are asked questions for each experiment. Take the time to explain fully everything that is asked of you and justify your answers.
It is natural that you will need to ask questions regarding this assignment. Please ask them in the course forum.
Background on On/Off traffic flows
Theoretical statistics of on-off processes
The mean of an on-off process of this type is half the on-rate. This is because half the time it is generating traffic at this rate, and the other half of the time it is generating no traffic.
If the mean of one of these on-off processes is m, the standard deviation of this process will be
m . A combination of n of these on-off processes will therefore have a mean of n × m, and a standard deviation of m .
Netml simulation of on-off processes.
Netml simulates on-off processes, but it is left for you to decide how well the simulation honours the theoretical statistics.
Setup for single flows
Step 1: The network
Create a network similar to the following:-
Figure 1: Network to be set up
Step 2: NS3 properties of the links
Set up the NS3 properties of the links as follows:
Figure 2: NS3 properties of links
Please note the following:-
• The link names in the diagram correspond to the link names in the NS3 parameter setup.
• The Base Addresses are properly set up. (Hint: The bases can be set up automatically by simply ticking the box to the left of -BaseAddress-.)
• All links have a capacity of 100Mbps, except the bottleneck, which has a capacity of 200Kbit/sec. (Hint: In the settings box there must be no space between the -400- and the -kbps-, otherwise the software will ignore the -kbps-. Also the -k- in -kbps- must be in lower case.)
• Although the links shown in netml are one way links, they will be simulated as two way links in NS3.
• The buffer sizes have been set to 20 packets.
Step 3: Traces and plots
Node A
Set up a trace and a plot sourced from node A for the enqueing of packets onto the -sent- link as shown in the following diagrams. Traffic monitored here is the aggregated traffic that has not entered the bottleneck yet and there is no loss.
Figure 3: Trace sourced from node A
Figure 4: Plot sourced from node A
Node C
Set up a trace and a plot sourced from node C for the enqueing of packets onto the -received- link.
Traffic monitored here has passed through the bottleneck and does not include dropped traffic.
Experiments
Experiment 1: UDP flow of 100 Kbit/sec
You are required to set up a flow of 150 Kbit per second. Plot the sent traffic and the received traffic.
Questions
1. The plots (10 marks )
2. What are the statistics of this traffic? (5 marks )
3. How much loss would you expect and how much did you get? (10 marks )
Experiment 2: UDP flow of 400 Kbit/sec
You are required to set up a flow of 400 Kbit per second. Plot the sent traffic and the received traffic.
Figure 5: Constant UDP traffic
The above image depicts the parameter settings for 8Mbit per second. Explanation:• The interval between sending packets is 0.001.
• The size of each packet is 1000 bytes.
• There are 8 bits per byte.
Questions
1. The plots (10 marks )
2. Explain the scale in the plot of the received traffic. (5 marks) (Hint: You will need to do calculations, so it is very important to get this right.)
3. Are the numbers calculated from the graph correct given the capacity of the link? (3 marks )
4. How much loss would you expect and how much did you get? (10 marks)
Experiment 3: On/Off UDP flow of 520 Kbit/sec
You are required to set up On/Off UDP traffic with a peak flow of 520 Kbit per second.
Use the following settings as shown in Figure 6:-
• Set the ON flow rate to 520 Kbps. (Due to a glitch in Netml, 520 Kbps is the only flow rate that is possible.)
• Set the on and off intervals to 0.3 seconds.
• Set the on and off distributions to exponential.
• Set the transport protocol to UDP.
Figure 6: Setting parameters for On Off UDP traffic
Plot the sent traffic and the received traffic.
Questions
1. The plots (10 marks )
2. From the theoretical statistics of on-off traffic, calculate what the mean of the generated traffic should be. (6 marks )
3. Calculate the area under sent traffic plot. From this, estimate the actual mean of the generated traffic. (5 marks )
4. Compare the actual mean of the generated traffic with your calculation in 1. (3 marks )
5. Given the mean of the generated traffic estimated from the graph, use the theory of on-off traffic to calculate the expected standard deviation. (4 marks )
6. Use the normal loss formula to calculate the expected loss using the actual mean and the standard deviation calculated in 4. (7 marks )
7. Calculate the area under the received traffic. From this, estimate the mean of the received traffic. (5 marks )
8. Deduce the loss from the difference between the areas of the sent and received traffic. (2 marks)
9. Compare the graphically obtained loss in 7 with your manually calculated answers in 5.
Have an attempt to explain descrepancies. (4 marks )
Experiment 4: On/Off UDP flow of 520 Kbit/sec with tiny buffers
Experiment 2 used a buffer size of 20 packets. Reduce the buffer size on the bottleneck link to just 2 packets, and repeat Experiment 3.
Questions
1. The plots. (10 marks )
2. Look at the shape of the received traffic in Experiment 3, and Experiment 4. What effect does reducing the buffer size have on the shape of the throughput? Why? (8 marks )
3. Deduce the loss from the graphs as before. (8 marks )
4. Does the loss with the smaller buffer agree more closely with the loss calculated using the normal loss formula calculated in Experiment 3? Why should it agree better? (4 marks )
Experiment 5: Using TCP flows instead of UDP flows
Leaving the buffer size on the bottleneck link as per Experiment 4, change the flow from a UDP flow to a TCP flow.
Questions
1. The plots. (10 marks )
2. Compare the results with that of Experiment 4. Is the sent traffic different from the sent traffic of experiment 4? What is different? Why is it different? (6 marks )
3. Deduce the loss from the graphs as before. (8 marks )
4. How does the loss compare to that in Experiment 4? Explain the difference. Why is this happening? (6 marks )
Experiment 6: On/Off UDP flow with 300 Kbps capacity link
Use UDP flow. Use a link size of 300 Kbps and buffer size of 2 packets for the bottleneck link. Estimate the area under both plots.
Questions
1. The plots (10 marks )
2. Calculate the mean of the generated traffic. Calculate the mean of the received traffic.
Calculate the loss. (10 marks )
Experiment 7: Aggregation of traffic flows
Increase the capacity of the bottleneck link to 3 Mbps. Change the flow from being a TCP flow back to being a UDP flow. Replicate this flow 9 times making a total of 10 such flows.
(Hint: Right click on the flow and select -copy-. Then go to the edit menu and use -paste multiple-.)
Figure 7: Configuration for Experiment 7
Questions
1. The plots (10 marks )
2. Calculate the mean of the generated traffic. Calculate the mean of the received traffic. Calculate the loss. (10 marks )
3. Compare the loss in this experiment with that is Experiment 6. Given that mean of the generated traffic and the mean of the link capacity has been multiplied by 10, why hasn't the loss been multiplied by 10? (4 marks)
Submission
Your submission will take the form of a single PDF file.
Present the results for each experiment in order, each of which will have the plots followed by the answers to the questions in order.
For each experiment
• Provide a heading for the experiment. Use the experiment number and experiment title.
• For each question
? Provide a heading. Use the question number and an abreviated title.
? Answer the question.
? The plots should have the size reduced, so that the two plots can appear side by side on the page. Do not include the entire HTML page generated by the simulations.
GET ANSWERS / LIVE CHAT