Scheduling Criteria

Objectives

1. Understand criteria for evaluating scheduling algorithms

Notes

• This is 5.2
• We have devised multiple algorithms for scheduling over the years.
• It is reasonable we have established criteria to judge scheduling algorithms
• Outcome 6: "Compare and critique preemptive and non-preemptive scheduling policies."
• The book suggests 5 criteria
  1. CPU utilization
     • Just what it says, how busy is the CPU kept.
  2. Throughput
     • processes completed per time unit.
     • An average
  3. Turnaround time
     • Time from the start of a process until it is finished.
     • An average.
  4. Waiting time
     • How much time is a ready process waiting for CPU access?
  5. Response Time.
     • Time from submission until the first response is produced.
     • More of a measure of how long after I start a process does it start doing something.
     • Not total wait time, but beginning wait time.
• These are different from what I was taught, but mine included priority queues, which we will see later.
• We generally want low averages for all of these.
  • Minimal average wait time.
  • Minimal average turnaround time.
• And we probably want to maximize
  • CPU utilization
  • throughput
• But the book suggests that minimizing maximum response time could be a goal.
• They suggest that minimizing the variance in response time is the important measure for interactive systems.
  • IE users get consistent results, even if they are slower.
• In the real world
  • Many Processes
  • Hundreds of bursts each, both CPU and I/O
• For the examples:
  • Few processes
  • One CPU burst each.
  • And they will focus on average waiting time.
• For the first pass, we will consider algorithms with a single CPU and a single queue.