Ch'05 outline

1
Ch.05 outline

• Basic concepts
• CPU-I/O burst cycle
• CPU scheduler
• Preemptive scheduling
• Cost
• dispatcher

2
Ch.05 outline (cont)

• Scheduling criteria
• CPU utilization, throughput, turnaround time,
  waiting time, response time
• Scheduling algorithms
• FCFS, nonpreemptive
• SJF, estimation of next CPU burst
• Exponential average
• SRTF, preemptive

3
Ch.05 outline (cont)

• Scheduling algorithms (cont)
• Priority
• Starvation
• aging
• Round-robin
• Time slice (quantum), size
• Extreme cases

4
Ch.05 outline (cont)

• Scheduling algorithms (cont)
• Multilevel queue
• Multilevel feedback queue
• characteristics
• Multiple-processor scheduling
• Symmetric or asymmetric
• Real-time scheduling

5
Ch.05 outline (cont)

• Algorithm evaluation
• Deterministic or queueing models
• simulation
• Process scheduling models
• Solaris 2, Windows 2000, Linux

6
PRESENTATION

• Scheduling algorithms(?5.3)
• FCFS, nonpreemptive
• SJF, estimation of next CPU burst
• Priority
• Round-robin
• Multilevel Queue
• pp. 154 - 165

7
CPU scheduling

• Problems in preemption
• Preemption incurs costs because states must be
  saved restored (context switching)
• Processes sharing data need a new mechanism (Ch.
  7)
• Kernel-execution model
• No preemption for kernel processes due to
  event occurrences (delay)
• preserves data consistency but not good for RT
  or MP scheduling (sects., 4 5)

8
CPU scheduling

• Interrupts, main source of preemption
• interrupts preempt all processes, including
  lower priority interrupt process
• Atomic execution
• nonpreemption applied to a smaller unit
• portions of codes that must be executed as a
  single unit
• disabling interrupts is a solution but can
  cause overhead in MP systems

9
Scheduling algorithms

• FCFS (no priority, nonpreemptive)
• SJF
• Priority
• RR
• Multi-level queue
• Multi-level queue with feedback

10
MP scheduling

• Homogeneous vs heterogeneous
• fixed assignment
• load sharing (separate vs common que)
• careful programming needed if two
• processors access a common DB
• asymmetric MP
• master-slave relationship

11
RT scheduling

• Hard
• resource reservation
• Soft
• 1. higher priorities no degradation over
  time for RT processes
• 2. low dispatch latency
• preemption points for syscalls
• preemption of syscalls (sync in Ch. 7)

12
homework

• What is the meaning of the last two sentences of
  the first paragraph in page 155?
• Explain what the author is trying to say in the
  fourth paragraph (p.171).
• What is the basic limitation of algorithms based
  on execution time? Suggest a method of prediction
  for next CPU burst?

13
homework

• 4. What are the things you have to consider in
  determining the size of time quantum (time slice)
  in RR scheduling?