Chapter 3

CPU Scheduling
Outline
 Introduction to scheduling
 Process behavior
 When to schedule
 Categories of scheduling algorithms
 Scheduling algorithms goals (criteria)
 Scheduling algorithms
 Scheduling in batch systems
 Scheduling in interactive systems
 Scheduling in Real time systems
CPU scheduling

 The aim of process scheduling is to assign processes to be executed by

the processor or processors over time in a way that meets the system
objectives.
 Fundamentally, scheduling is a matter of managing queues to minimize
delays and to optimize performance.
 Scheduling is the basis of multi-programmed operating system.
 Scheduling in batch system was so simple. But in time sharing system,
algorithms are becoming more and more complex.
 Some main frame system- use both ways
 The scheduler decides whether a batch job or interactive user at the terminal should
go next.
CPU scheduling

 CPU was a scarce resources on these machines. Hence, a good

algorithm can make a huge difference.
 But with the advent of personal computer
 One process at a time
 CPU is a rarely scarce resource – PC are limited by user input and
application type
 The situation is different when it comes to high-end networked workstations and
servers. – sending e-mail and closing windows
 The scheduler has to worry about making efficient use of CPU because
switching is expensive.
Process Behavior

 CPU bound – long CPU burst

 I/O bound – long I/O burst

Waiting for I/O

 The key factor is CPU burst.

 As the CPU gets faster processes tend to get more I/O.
 And CPU is improving in much faster way than disks.
When to schedule

 Key issues to make scheduling decisions

 When a new processes are created.
 When a process switches from running state to waiting state.
 I/O or an invocation of wait() for the termination of one of the child process.
 When a process switches from running state to the ready state.
 When an interrupt occurs.
 When a process switches from waiting to ready state.
 Completion of I/O.
 When a process terminates.
When to schedule

 If a hardware provides periodic interrupts, scheduling

decision can be made at each clock interrupts.
 Non-Preemptive and preemptive
Under non-preemptive scheduling, once the CPU has
been allocated to a process, the process keeps the CPU
until it releases the CPU either by terminating or by
switching to the waiting state. Otherwise, it is
preemptive.
Categories of scheduling algorithms

 In different environment different algorithms are implemented

 Batch systems
 No user waiting impatiently. Hence, Non-preemptive or long time run preemptive are
acceptable
 Interactive systems
 Need preemption
 Most are general purpose – even malicious software can run.
 Real-Time systems
 Preemption is oddly enough, some times not needed
 Run only programs that are needed for further application at hand
 CPU scheduling performance metrics

 Processor Utilization  The ratio of busy time of the processor to the

total time passes for processes to finish.
Processor Utilization = (Processor buy time) / (Processor busy time + Processor idle time)
 Throughput  Number of processes that complete their execution per
time unit.
Throughput = (Number of processes completed) / (Time Unit)
 Turnaround time  Amount of time to execute a particular process.
tat = t(process completed) – t(process submitted)
CPU scheduling performance metrics

 Waiting time  Amount of time a process has been waiting in the ready
queue.
 Burst Time  Time of execution, when process is in Processor.
 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).
rt = t(first response) – t(submission of request)
Scheduling algorithms Goals

 All systems
 Fairness is important
 Comparable processes should get fair share
 Enforcing systems policy
 Safety control and payroll applications in a nuclear reactor
 Keeping all parts of the computer busy
 I/O bound and CPU- bound can be run together.
Scheduling algorithms Goals…

 Batch systems
 Throughput - maximum jobs per hr.
 Turn around time – minimize time between submission and termination
 CPU utilization – keep the CPU busy all the time

 Interactive systems
 Response time – respond to request quickly
 Proportionality –meet users expectations.
 Scheduling algorithms Goals…

 Real-Time
 Meeting deadlines – avoid losing of data
 Predictability – avoid quality degradation in multi-media systems
 Scheduling in Batch systems

 FCFS
 SJF
 Short time remaining next
 First Come First Served

 The process that requests the CPU first is allocated the CPU first.
 It’s the simplest of all algorithms
 It’s non-preemptive
 A blocked process come to an end of the queue when it’s ready to be
executed
 Easy to understand and to program
 A single linked list keeps track of all ready state processes
 The greater disadvantage is when executing I/O bound processes.
 First Come First Served

Process Burst Time (milliseconds) Arrival time

P1 24 0
P2 3 0
P3 3 0
 Suppose that the processes arrive in the order: P1 , P2 , P3
 The Gantt Chart for the schedule is:
P1 P2 P3

0 24 27 30
 Waiting time for P1 = 0; P2 = 24; P3 = 27
 Average waiting time: (0 + 24 + 27) / 3 = 17 (milliseconds)
 Average turn-around time: (24 + 27 + 30)/3 = 27
 Shortest Job First

 Another non-preemptive and the run time of the processes are

known in advance.
 When multiple batch jobs are sitting in a queue with the same
priority, the scheduler runs the shortest job first.
 Shortest Job First…

 SJF is optimal only if all jobs are available simultaneously.

 Consider five jobs whose running time is 2, 4, 1, 1, 1 respectively and
their arrival time is 0, 0, 2, 3, 3 respectively.
 Using SJF the average waiting time is 1.2
 Running in order of, A, C, D, E,B results.
 Short Job First…

Process Arrival Time Burst Time

P1 0.0 6
P2 0.0 4
P3 0.0 1
P4 0.0 5
 SJF (non-preemptive, simultaneous arrival)

P3 P2 P4 P1

0 1 5 10 16
 Average waiting time = (0 + 1 + 5 + 10)/4 = 4
 Average turn-around time = (1 + 5 + 10 + 16)/4 = 8
 Non-Pre-emptive SJF

Process Arrival Time Burst Time

P1 0.0 7
P2 2.0 4
P3 4.0 1
P4 5.0 4
 SJF (non-preemptive)

P1 P3 P2 P4

0 3 7 8 12 16

 Average waiting time(to start execution) = (0 + 6 + 3 + 7)/4 = 4

 Pre-emptive SJF

 Preemptive version of SJF.

 The scheduler always picks the processes whose remaining running time
is the shortest.
 Again the run time should be known in advance.
 The shortest new job has a great advantage over all processes.
 Pre-emptive SJF = Short remaining time first

Process Arrival Time Burst Time

P1 0.0 7
P2 2.0 4
P3 4.0 1
P4 5.0 4
 SJF (preemptive)
P1 P2 P3 P2 P4 P1

0 2 4 5 7 11 16

 Average waiting time = (9 + 1 + 0 +2)/4 = 3

Shortest Remaining Time First

Process ID Arrival Time Burst Time Turn Waiting Response

Around Time Time
Time

1 0 8 20 12 0

2 1 4 9 5 0

3 2 2 2 0 0

4 3 1 2 1 1

5 4 3 9 6 1

6 5 2 2 0 0
Shortest Remaining Time First
 Scheduling in interactive systems

Scheduling in interactive systems

 Round robin scheduling
 Priority scheduling
 Multiple queuing
 Shortest process next
 Guaranteed scheduling
 Round- Robin scheduling

 Processes will be given equal priority

 It uses quantum- slice of time
 Therefore each process will be given an interval of time (quanta of time).
 1 quanta, 2 quanta…..
 Processes are allowed to run in their given slice of time.
 The main issue in round- robin is time taken by context switching. E.g. 1msec. For
process switching and 4msec as a quanta
 Round- Robin…

 It’s better to set one quantum enough amount of time say 100msec to
improve the efficiency.
 But here what if ten users press return button at the same time?
 The two main things as a summary
 Setting the quantum two short causes to many process switches that will result in
performance degradation.
 But setting it too long may cause poor response to short interactive commands.
 Round-Robin….

Process Burst Time Q=20

P1 53
P2 17
P3 68
P4 24
P1 P2 P3 P4 P1 P3 P4 P1 P3 P3

0 20 37 57 77 97 117 121 134 154 162

 Typically, higher average turnaround than SJF, but better

response.
 Priority scheduling

 Each process is assigned a priority.

 The process with the highest priority is allowed to run.
 This idea is critical in a multi-user environment and also in a single user
with several applications running at a time.
 Priority will decrease at each clock tick (clock interrupt)
 When the priority level lower from that of the next highest one, the
process switching occurs.
 Plus to the above each priority will be given maximum allowed
quantum of time.
 Priority scheduling…

 Unix system has nice which allows the user to voluntarily reduce its
priority level.
 Process priority can be given statically and dynamically
 Statically-Role and privilege assignment in MIS systems
 Dynamically- I/O bound processes should be given the highest priority after waiting
for I/O response and asks to run again.
 This avoid unnecessary memory hold time
 A simple algorithm for giving this service is to set the priority to 1/f where f is fraction of
the last quantum that the process used.
 For e.g a process that uses 1 msec of its 50 msec quantum then it will be given priority
level equivalent to its quantum.
 Priority scheduling….

 It’s often convenient to create priority class and use round-robin in

each classes.
Priority scheduling….

Process Burst Time Priority

P1 10 3
P2 1 1
P3 2 4
P4 1 5
P5 5 2
 Multiple Queues

 Processes in the highest priority class will run for 1 quantum and
processes in the next highest class will run for 2 quanta processes in the
next class will run for 4 quanta and so on.
 Consider how a process that needs100 quanta to continuously run and finish its
computation. It would only need 7 swaps unlike 100 swaps in pure round- robin.
Multiple Queues….. Priority level examples
 Multiple queues….

 As a processes sank deep and deeper in to the priority queues it would

be run less and less frequently.
 This saves the CPU for short and interactive processes.
 A policy of making a process that was decided to run for a long time to
become interactive later.
 E.g. copying a file or installing a memory demanding software
 Guaranteed Scheduling

 Make real promises to the users about performance and live up to

them.
 If n users logged in each will get 1/n of CPU power.
 If a user have n processes to run each will get 1/n of the CPU cycle.
 To make good on the promise, The system must keep track of how
much CPU the process has had since creation.
 i.e. ratio of actual CPU time consumed to CPU entitled.
 0.5, 2.0 ……
 The algorithm then runs the process with the lowest ratio.
 Real-Time Scheduling

 Hard real-time systems – required to complete a critical task within a guaranteed

amount of time.
 Resource reservation – knows how much time it requires and will be scheduled if it
can be guaranteed that amount of time
 Requires special purpose software running on hardware dedicated to their critical
process
 Soft real-time computing – requires that critical processes receive priority over less
fortunate ones.
 System must have priority scheduling where real time processes are given the
highest priority (no aging to keep priorities constant)
 Dispatch latency must be small – need to be able to preempt system calls
particularly in long, complex calls at safe points.
Reading Assignment

 Algorithm evaluation
 Scheduling in different OS
windows
Solaris
Unix
Mobile OS
*******End of Chapter 3*************