CS 3733 Operating Systems, Fall 2010 Assignment 3 Comments

This assignment was graded out of 30 points.

• Most people had little trouble with doing most of this assignment.
• Most people had trouble with the answer to the question for Part 2.
  • What is meant by close enough to zero?
  • You can only determine whether something is close to zero in relationship to another number.
  • Example:
    • I estimate the diameter of the earth at its poles to be 7900 miles. Is this a close enough estimate? The correct answers is about 7901 miles. The difference is 1 mile which is sufficiently small for most uses.
    • I estimate the distance between my office and this classroom is 1 mile. This estimate is within 1 mile of the correct answer. Is this a good estimate?
  • Here are two different analyses for FCFS:
    • The correct answer from the simulator is 5739.3, and the estimate is 5821.53. The difference is 82.23, or about 1.4% off. This seems like a good estimate.
    • The context switch time changes the AWT by 447, according to the simulator. The estimate says this should be 529.23, a difference of 82.23, or 18.4%. Not too good an estimate.
• The hardest part was answering the question in Part 3.
  • You were supposed to realize that the context switch time affects the average waiting time because of two factors: the number of context switches (obviously) and the load average.
  • While PSJF might have more context switches than SJF, it is also possible that it will have a smaller load average.
  • In this case it is possible that the product of the load average and the number of context switches will be smaller for PSJF than SJF.
  • Note that the result of the one experiment that you did cannot prove that something is always true.
  • However, if you did a different experiment, it would be possible to show that the statement was not always true.