Scheduling: Introduction

Chapter 7 of Operating System: Three Easy Pieces

Operating Systems by Dr. Praveen Kumar @ CSED, VNIT Nagpur 1

Review: Process State Transitions

 How to transition? (“mechanism”) – Previous lectures

 When to transition? (“policy”) – on context switch OS scheduler decides

which process to run next, from set of ready processes

Descheduled
Running Ready
Scheduled

I/O: initiate I/O: done

Blocked

Operating Systems by Dr. Praveen Kumar @ CSED, VNIT Nagpur 2

Exercise

 Think of examples in life where scheduling is involved

Operating Systems by Dr. Praveen Kumar @ CSED, VNIT Nagpur 3

Scheduling: Outline

 How should we develop a basic framework for thinking about

scheduling policies?

 What metrics are important?

 What are the key assumptions?

 What basic approaches have been used in different computer systems

Operating Systems by Dr. Praveen Kumar @ CSED, VNIT Nagpur 4

Scheduling: Terminology

 Workload: set of job descriptions (arrival time, run time)

 Job: current CPU burst (CPU time used in one stretch) of a process

 Performance Metrics: measurement of scheduling quality

 Turnaround time: The amount of time taken to execute the particular job

Turnaround time = (completion time - arrival time)

 Waiting Time: The amount of time a process has been waiting in the ready
queue. Waiting Time = (turnaround time - burst time)

 Response time: The amount of time from when a request was submitted
until the first response is produced. Response Time = (the time at which a
process get the CPU first time - arrival time)

 Throughput: How many jobs complete per unit of time

 Fairness: All jobs get same amount of CPU over some time interval
Operating Systems by Dr. Praveen Kumar @ CSED, VNIT Nagpur 5
Exercise

 Take a real life scenario/example and try to explain these metrics in

that context.

Operating Systems by Dr. Praveen Kumar @ CSED, VNIT Nagpur 6

Scheduling: A beginning

 Workload assumptions:
1. Each job runs for the same amount of time.

2. All jobs arrive at the same time.

3. All jobs only use the CPU (i.e., they perform no I/O).

4. The run-time of each job is known.

 Don’t worry… we will relax them as we move along

Operating Systems by Dr. Praveen Kumar @ CSED, VNIT Nagpur 7

Scheduling Metrics

 Performance metric: Turnaround time

 The time at which the job completes minus the time at which the job
arrived in the system.

𝑻𝒕𝒖𝒓𝒏𝒂𝒓𝒐𝒖𝒏𝒅 = 𝑻𝒄𝒐𝒎𝒑𝒍𝒆𝒕𝒊𝒐𝒏 − 𝑻𝒂𝒓𝒓𝒊𝒗𝒂𝒍

Operating Systems by Dr. Praveen Kumar @ CSED, VNIT Nagpur 8

First In, First Out (FIFO)

 First Come, First Served (FCFS)

 Very simple and easy to implement

 Example: three jobs arrive at t=0 in the order A,B,C

 Each job runs for 10 seconds.

 Gantt chart below : Illustrates how jobs are scheduled over time on a CPU
A B C

0 20 40 60 80 100 120

Time (Second)

𝟏𝟎 + 𝟐𝟎 + 𝟑𝟎
𝑨𝒗𝒆𝒓𝒂𝒈𝒆 𝒕𝒖𝒓𝒏𝒂𝒓𝒐𝒖𝒏𝒅 𝒕𝒊𝒎𝒆 = = 𝟐𝟎 𝒔𝒆𝒄
𝟑

Operating Systems by Dr. Praveen Kumar @ CSED, VNIT Nagpur 9

Why FIFO is not that great?

 Let’s relax assumption 1: Each job no longer runs for the same
amount of time.

 Example: three jobs arrive at t=0 in the order A,B,C

 A runs for 100 seconds, B and C run for 10 each.

 Draw the Gantt chart and calc. Avg. turnaround time

A B C

0 20 40 60 80 100 120

Time (Second)

𝟏𝟎𝟎 + 𝟏𝟏𝟎 + 𝟏𝟐𝟎

𝑨𝒗𝒆𝒓𝒂𝒈𝒆 𝒕𝒖𝒓𝒏𝒂𝒓𝒐𝒖𝒏𝒅 𝒕𝒊𝒎𝒆 = = 𝟏𝟏𝟎 𝒔𝒆𝒄
𝟑

Operating Systems by Dr. Praveen Kumar @ CSED, VNIT Nagpur 10

FIFO problem: Convoy Effect

 Think of a similar scenario you may have encountered personally

 What do you do in that situation?

Operating Systems by Dr. Praveen Kumar @ CSED, VNIT Nagpur 11

Shortest Job First (SJF)

 Run the shortest job first, then the next shortest, and so on
 Non-preemptive scheduler

 Same example: three jobs arrive at t=0 in the order A,B,C

 A runs for 100 seconds, B and C run for 10 each.

 Provably optimal when all processes arrive together

B C A

0 20 40 60 80 100 120

Time (Second)

𝟏𝟎 + 𝟐𝟎 + 𝟏𝟐𝟎
𝑨𝒗𝒆𝒓𝒂𝒈𝒆 𝒕𝒖𝒓𝒏𝒂𝒓𝒐𝒖𝒏𝒅 𝒕𝒊𝒎𝒆 = = 𝟓𝟎 𝒔𝒆𝒄
𝟑

Operating Systems by Dr. Praveen Kumar @ CSED, VNIT Nagpur 12

SJF with Late Arrivals from B and C

 Let’s relax assumption 2: Jobs can arrive at any time.

 Example: A arrives at t=0 and needs to run for 100 seconds.

 B and C arrive at t=10 and each need to run for 10 seconds

 SJF is non- preemptive, so short jobs can still get stuck behind long ones

[B,C arrive]
A B C

0 20 40 60 80 100 120

Time (Second)

𝟏𝟎𝟎 + 𝟏𝟏𝟎 − 𝟏𝟎 + (𝟏𝟐𝟎 − 𝟏𝟎)

𝑨𝒗𝒆𝒓𝒂𝒈𝒆 𝒕𝒖𝒓𝒏𝒂𝒓𝒐𝒖𝒏𝒅 𝒕𝒊𝒎𝒆 = = 𝟏𝟎𝟑. 𝟑𝟑 𝒔𝒆𝒄
𝟑

Operating Systems by Dr. Praveen Kumar @ CSED, VNIT Nagpur 13

Shortest Time-to-Completion First (STCF)

 Add preemption to SJF

 Also knows as Preemptive Shortest Job First (PSJF)

 When a new job enters the system:

 Preempts running task if its remaining CPU burst is more than that of new
arrival

Operating Systems by Dr. Praveen Kumar @ CSED, VNIT Nagpur 14

Shortest Time-to-Completion First (STCF)

 Example:
 A arrives at t=0 and needs to run for 100 seconds.

 B and C arrive at t=10 and each need to run for 10 seconds

 Draw Gantt chart and calc. avg. TT

[B,C arrive]
A B C A

0 20 40 60 80 100 120

Time (Second)

(𝟏𝟐𝟎 − 𝟎) + 𝟐𝟎 − 𝟏𝟎 + (𝟑𝟎 − 𝟏𝟎)

𝑨𝒗𝒆𝒓𝒂𝒈𝒆 𝒕𝒖𝒓𝒏𝒂𝒓𝒐𝒖𝒏𝒅 𝒕𝒊𝒎𝒆 = = 𝟓𝟎 𝒔𝒆𝒄
𝟑

Operating Systems by Dr. Praveen Kumar @ CSED, VNIT Nagpur 15

New scheduling metric

 Given our new assumption, STCF is provably optimal. Hence used in

early batch processing computing systems

 With advent of Time sharing demand for interactive performance

came

 Response time: The time from when the job arrives to the first time
it is scheduled.

𝑻𝒓𝒆𝒔𝒑𝒐𝒏𝒔𝒆 = 𝑻𝒇𝒊𝒓𝒔𝒕𝒓𝒖𝒏 − 𝑻𝒂𝒓𝒓𝒊𝒗𝒂𝒍

 STCF and related disciplines are not particularly good for response time.

How can we build a scheduler that is

sensitive to response time?

Operating Systems by Dr. Praveen Kumar @ CSED, VNIT Nagpur 16

Round Robin (RR) Scheduling

 Time slicing Scheduling

 Run a job for a time slice and then switch to the next job in the run
queue until the jobs are finished.
 Time slice is sometimes called a scheduling quantum.

 It repeatedly does so until the jobs are finished.

 The length of a time slice must be a multiple of the timer-interrupt period.

RR is fair, but performs poorly on metrics

such as turnaround time

Performance and fairness are often at odds in scheduling

Operating Systems by Dr. Praveen Kumar @ CSED, VNIT Nagpur 17

RR Scheduling Example

 A, B and C arrive at the same time.

 They each wish to run for 5 seconds.

A B C
0 + 5 + 10
𝑇𝑎𝑣𝑒𝑟𝑎𝑔𝑒 𝑟𝑒𝑠𝑝𝑜𝑛𝑠𝑒 = = 5𝑠𝑒𝑐
3

0 5 10 15 20 25 30
Time (Second)

SJF (Bad for Response Time)

A B CA B CA B CA B CA B C

0+1+2
𝑇𝑎𝑣𝑒𝑟𝑎𝑔𝑒 𝑟𝑒𝑠𝑝𝑜𝑛𝑠𝑒 = = 1𝑠𝑒𝑐
3

0 5 10 15 20 25 30
Time (Second)

RR with a time-slice of 1sec (Good for Response Time)

Operating Systems by Dr. Praveen Kumar @ CSED, VNIT Nagpur 18

The length of the time slice is critical.

 The shorter time slice

 Better response time

 The cost of context switching will dominate overall performance.

 The longer time slice

 Amortize the cost of switching

 Worse response time

Deciding on the length of the time slice presents

a trade-off to a system designer

Operating Systems by Dr. Praveen Kumar @ CSED, VNIT Nagpur 19

Incorporating I/O

 Let’s relax assumption 3: All programs perform I/O

 Example:
 A and B need 50ms of CPU time each.

 A runs for 10ms and then issues an I/O request

 I/Os each take 10ms

 B simply uses the CPU for 50ms and performs no I/O

 The scheduler runs A first, then B after

Operating Systems by Dr. Praveen Kumar @ CSED, VNIT Nagpur 20

Incorporating I/O (Cont.) with STCF

A A A A A B B B B B

CPU

Disk

0 20 40 60 80 100 120 140

Time (msec)

Poor Use of Resources

A B A B A B A B A B Maximize the
CPU utilization by
CPU treating each 10ms
sub job of A as new
Disk job

0 20 40 60 80 100 120 140

Time (msec)

Operating Systems by Dr. Praveen Kumar @ CSED, VNIT Nagpur 21

Incorporating I/O (Cont.)

 When a job initiates an I/O request.

 The job is blocked waiting for I/O completion.

 The scheduler should schedule another job on the CPU.

 When the I/O completes

 An interrupt is raised.

 The OS moves the process from blocked back to the ready state.

Operating Systems by Dr. Praveen Kumar @ CSED, VNIT Nagpur 22