This being the first lab exercise, I would like to make sure that you understand the “protocol” very well.
Deadline: Friday, Feb 1, 5 pm. Be sure that you have completed the exercise well before the deadline. I will not tolerate excuses (power fail, server down, etc.). In our context, these things happen, and will happen on the day the exercise is due for submission. I will give you a grace period of 1 week (7 calendar days), provided you are willing to be graded out of 50% of the marks assigned to the exercise.
You may do the project in two
After you done with the exercise, prepare a folder of the following docs:
An algorithm
The complete program
Input, output data (in the form of files)
Error reports, if any.
A brief 1 page report which brings out the comparison, evaluation or analysis of what ever it is that you are trying to study.
You are then required to submit the folder to your TA as follows:
If the sum of the last digit in the entry no. of all members in the group is EVEN then send it to P. Ambica (csa00006@cse.iitd.ernet.in). Otherwise, send it to Amit Jain (mailto:csa00011@cse.iitd.ernet.in).
The exercise itself consists of the following:
You are required to simulate queuing behavior, and determine whether an M/M/1 analysis is adequate and acceptable. We will study two different queuing systems:
The first is a queuing system that comes very close to the assumption that we have made for M/M/1 queue. That is assume:
(a) Messages arrive at a transmitter as per Poisson process (you will use an available uniformly distributed pseudo random no. generator to generate arrivals that are exponentially distributed. Each message is exponentially distributed in length.
(b) The transmitter has a fixed channel capacity.
(c) There is a single queue, with unlimited number of buffers.
You will simulate this system for a certain duration (figure out how long should the simulation must run). And while the simulation progresses, make sure you keep track of (a) which message arrives and when so that you can figure out the time spent by each message, (b) the length of the queue as a function of time so that you can compute the averages.
Having compared the average waiting plus service time with those that you would obtain from theory, see what happens when the number of buffer is finite (16, 8, 4, 2).
The above was to make sure that you understand how queues are simulated.
Now you will simulate the real thing: FOUR sources generate data at some peak rate rate (and no more), but the average is much lower than the peak. Assume that the inter-arrival time for each of them is different, but uniform distributed (say 0 and T_i). The packets are fixed length. And so is the channel capacity. Further, that there is a maximum no. of buffers. NOW see if the M/M/1 assumption if applied gives results that approximate the actual behavior.
GOOD LUCK.