We observe a closed system for 30 minutes, during which 1600 tasks are completed, from 12 terminals. Each terminal (source of tasks) has a think time of 12s between receiving the response to one task, and submitting the next task. The system has a CPU and two disks, one is fast, the other is slow. The fast disk can deliver a block in 1/2 of the time that the slow disk would take. During the observation period, 32000 accesses occur at the fast disk, and 12000 accesses at the slow disk. During the observation period, the CPU is busy for 1080s, the fast disk is busy for 400s, and the slow disk is busy for 600s.

a) What is the response time for jobs in the observed system?

b) As a function of N, the number of terminals, what are the high-load bounds for throughput and response time?

c) As a function of N, what are the low-load bounds for throughput and response time?

d) How many terminals can we support, while keeping the response time below 3 s?

e) Suppose we plan to replace the CPU with a CPU whose speed is double (thus the time to execute each job would decrease so it becomes 1/2 of the current time). Estimate the number of terminals we can support, with response time below 3s.

f) Propose a design that can support 45 terminals with response time below 3s.