Operating Systems- Process (CPU)

schedulingSeehwan YooMSE.cis.dku

Review

• Computer systems (hw)• Main components: CPU, mem,

I/O devices• Inter-connect: bus

• I/O device• interaction w/ external world• Comm. w/ cpu

• Interrupt• Polling

• Direct memory access

• SW in computers• OS kernel, user applications• Dual-mode execution

• For protection• System call – interface to OS

• Process• Abstraction of proces-

sor (CPU) / machine• Execution context• Cf.) program

• Memory layout• Address space

• Process states• State transition

• Time-sharing system• Implementation with

timer interrupt

Process scheduling

• Determine which process to run• Kernel maintains queues

• Data structure!• Ready Q (runq)

• Store processes in ready state• Device Q (ioq)

• Store processes in waiting state (per-device)

• Scheduler picks one of process to run during the next time slot

• How?

Overall picture of schedul-ing

Scheduling

• Diverse policies• Fair-share• First-In-First-Out• Shortest-Job-First

• Good scheduler?• Evaluation criteria

• CPU utilization• Throughput• Turn-around time• Waiting time• Response time

Scheduling Criteria

• CPU utilization – keep the CPU as busy as possible• Throughput – # of processes that complete their ex-

ecution per time unit• Turn-around time – amount of time to complete a

particular process• Waiting time – amount of time a process has been

waiting in the ready queue• Response time – amount of time it takes from when

a request was submitted until the first response is produced, not output (for time-sharing environment)

Anti-break!

• Waiting time of blue-process?• Response time of red-process?• Turn around-time of blue-process at time 3?• Throughput of the system?• CPU utilization?

a b c d e f

First-In, First-Out

• This is not a fair-share scheduler!• First-comer run until it is completed, or

• Until it makes I/O request, or• Until it voluntarily yield CPU

FIFO (FCFS) exampleProcess Burst Time

P1 24

P2 3

P3 3

• Suppose that the processes arrive in the order: P1 , P2 , P3 (at time 0)

• The Gantt Chart for the schedule is:

• Waiting time for P1 = 0; P2 = 24; P3 = 27• Average waiting time: (0 + 24 + 27)/3 = 17

P P P1 2 3

0 24 3027

FIFO (FCFS) another exam-ple!

Suppose that the processes arrive in the order:

P2 , P3 , P1 • The Gantt chart for the schedule is:

• Waiting time for P1 = 6; P2 = 0; P3 = 3• Average waiting time: (6 + 0 + 3)/3 = 3• Much better than previous case• Convoy effect - short process behind long process

• Consider one CPU-bound and many I/O-bound processes

P 1

0 3 6 30

P 2 P 3

Then, How about SJF?

• Take the shortest job at first hand! ProcessArriva l TimeBurst Time

P1 0.0 6

P2 2.0 8

P3 4.0 7

P4 5.0 3

• SJF scheduling chart

• Average waiting time = (3 + 16 + 9 + 0) / 4 = 7

P 3

0 3 24

P 4 P 1

169

P 2

And How about SRTF?- Running scheduler at every time tick

• Take the remaining time job at first hand!• Now we add the concepts of varying arrival times and preemption to

the analysis ProcessA arri Arrival TimeT Burst Time

P1 0 8

P2 1 4

P3 2 9

P4 3 5• SRTF Gantt Chart

• Average waiting time = [(10-1)+(1-1)+(17-2)+5-3)]/4 = 26/4 = 6.5 msec

P 4

0 1 26

P 1 P 2

10

P 3P 1

5 17

Forget about it…

• Only exist in theory• Who knows the shortest job?

• It is the most well-known NP-hard problem• Homework) survey halting problem in the Internet

But, an approximation can be made• From the last execution history

• Commonly, α set to ½

:Define 4.

10 , 3.

burst CPU next the for value predicted 2.

burst CPU of length actual 1.

1n

thn nt

.1 1 nnn t

Like this,

We can put them all to-gether• Or stop it by now;• Break time

Fair-share schedulerRound-robin (RR)• Gives no priority to all process• Pick up one process in an equal probability

• If all the processes have the same time quantum, then all the pro-cesses would utilize CPU time evenly

• Implementation would be…• Randomly pick up a process in the run queue• When time quantum is expired, put it in the retired queue• When there is no process in the run queue, switch run queue with re-

tired queue

• In practice, • Pick the first process in the run queue• Run it until time quantum is expired• When time quantum is expired, put it in the run queue (tail)

RR property & example• Max (worst-case) response time?• Example Process Burst Time

P1 24

P2 3

P3 3 The Gantt chart is:

Typically, higher average turnaround than SJF, but better response q should be large compared to context switch time q usually 10ms to 100ms, context switch < 10 usec

P P P1 1 1

0 18 3026144 7 10 22

P 2 P 3 P 1 P 1 P 1

Priority-based scheduling

• No fair-share scheduler!• Give a priority to a process

• Scheduling according to the priority• Can be fixed• Can be dynamically changed

• Consider low-class citizens!• Low-priority process could be starved• Respect the seniors!

• Apply aging• Increase priority when a process is in the low priority for a long

time

Property of priority sched-uling• Deterministic response time

• Calculate it by hand!

• Used in real-time systems• Rate-monotonic• Earliest-Deadline-First (EDF)

Example of priority sched-uling

ProcessA arri Burst TimeT Priority

P1 10 3

P2 1 1

P3 2 4

P4 1 5

P5 5 2

• Priority scheduling Gantt Chart

• Average waiting time = 8.2 msec

1

0 1 19

P 1 P 2

16

P 4P 3

6 18

P

Re-think scheduling!

• Basic ideas• Fair-share• First-in-First-out (FIFO, or queue)• Real-time

• Do the right thing!• Does your application really want?• Compare

• file copy (doing a lot of I/O operation)• Hanoii (scientific application for complex calculation)

Thing to know – when you schedule process• Workload characteristics

• CPU-bound jobs (CPU-intensive)• I/O-bound jobs (I/O-intensive)

• Giving long CPU time for I/O-bound jobs• Giving high priority for CPU-bound jobs• What actually happens in the scheduler is

• Complex• Homework) survey the Linux’s Scheduler

A theory – Little’s Theorem

• n = average queue length• W = average waiting time in queue• λ = average arrival rate into queue• Little’s law – in steady state, processes leaving queue

must equal processes arriving, thus: n = λ x W

• Valid for any scheduling algorithm and arrival distribution

• For example, if on average 7 processes arrive per second, and normally 14 processes in queue, then average wait time per process = 2 seconds

Summary page

• Various schedulers• FIFO, SJF, SRTF, RR, priority-based

• Evaluation criteria• Throughput, …time(s)

• Workload characteristics• I/O or CPU-bound tasks

• Think about the scheduler operation when I/O is involved?