A: Determine the maximum sustained bandwidth (in bits per second) that can be achieved as a function of message size.
B: Determine the minimum round-trip latency as a function of message size.
Each of the workstations connected to the switch has 4 processors. Variations of these experiments include:
1. Load the CPU with independent processes which do not directly affect the network and determine how the network performance is affected.
2. Load the network by running multiple almost identical processes which communicate between the same pair of machines. In the case of bandwidth, we are interested in the total bandwidth of all connections.
3. Load the network with transmissions between random pairs of machines.
You will have to write a program to do the communication. The simplest way is to use sockets or TLI.
In addition, each experiment can be modified as follows:
X. Use UDP instead of TCP and measure the error rate, the fraction of messages that do not get through under various loadings. Add your own error handling to guarantee that the transmission is done correctly and compare the bandwidth or latency with TCP which whould be error free.
Choose one of the following 12 projects:
A1, A2, A3, B1, B2, B3, A1X, A2X, A3X, B1X, B2X, or B3X.
No two students will do the same project. You pick a project by
posting to utsa.cs.6553.d. I must approve your choice.
You should compare your ATM results to those obtained using our Ethernet.
You will produce two written reports and give a presentation in class. The first report will be a progress report containing details of the design of your experiment. The final report will be a complete report with all of the details including the code you used and the results. The presentation will be in class and last about 30 minutes.
Important dates:
First Report: Turned in to me by March 28 (sooner, if possible)
Final Report: Due April 11
Presentation: You must be prepared to give the presentation by April 16.