CS 3733 Operating Systems, Fall 1999 Assignment 2 Comments

Warning: This is not for the current semester.

1. Keep in mind that you cannot prove something is always true by doing a few (or even mnay) experiments.
2. Since time units are arbitrary in the simulator, it does not make sense to say that something like turnaround time is large. You need to compare it with something, usually the turnaround time of another run.
3. If you create a number of processes with the same uniform distribution of CPU bursts, all of the processes will have approximately the same characteristics. If the number of bursts is large, all of the processes will have some short bursts and some long bursts. You need to use two groups of processes if you want some processes to have all short bursts and some to have all long bursts.
4. You must use the same processes on all runs that you compare. Use same CPU burst distribution on all runs if you want to compare the affect of different algorithms.
5. The load average is the total waiting time divided by the total time. The simulation gives the average waiting time. To get the total waiting time, multiply by the number of processes.
6. What is meant by the expressions small quantum and large quantum? You need to compare it to something. A quantum is large if it is large compared to the longest CUP burst time. It is small if is is small compared to the smallest CPU burst time.
7. The goal is not to increase the CPU utilization. Users want to run on machines with low CPU utilization because this usually means better response. A CPU utilization of 68% is quite high.
8. If the quantum is greater than the largest CPU burst, the quantum does not expire RR behaves like FCFS. If the quantum is small, there is no reason to believe that RR will behave like SJF. There are some circumstances in which it will do better than FCFS, but often it will not.