Question

• Implement the above simulation for the 3 scheduling algorithms, FIFO, SJF, SRT.

• Assume a value for k, which is the time interval during which processes may arrive. Ex: k = 1000.

• Using a random number generator, derive n arrival times, Aᵢ, for all processes, distributed uniformly within the interval [0 : k].

• Choose an average total CPU time, d, and a standard deviation, v, and derive n total CPU times, Tᵢ, using a normal (Gaussian) distribution.

• Repeat the simulation for different values of d and v. (For simplicity, v could be just be a fixed percentage of d.)

• The values d and v should be selected relative to k. The value k/n represents the frequency of process arrivals. When d is much smaller than k/n, then most processes run in isolation, without having to compete with other processes for the CPU. As a result, all scheduling algorithms should produce similar results. On the other hand, when d is much larger than k/n, then many processes will overlap in time and compete for the CPU. Consequently, different scheduling algorithms should yield different results.

• To compare the different algorithms, plot the values d/ATT over d. The value d/ATT shows how competition for CPU slows down the processes. The smaller the value, the worse the overall performance. The ideal case is d/ATT = 1, which indicates that all processes ran without any delays.