See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/249645533

A Comparative Study of CPU Scheduling Algorithms

Article · July 2013

Source: arXiv

CITATIONS READS
49 16,657

2 authors, including:

Neetu Goel
Vivekananda Institute Of Professional Studies
4 PUBLICATIONS   72 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

CPU Scheduling View project

All content following this page was uploaded by Neetu Goel on 21 May 2015.

The user has requested enhancement of the downloaded file.

 Neetu Goel, R.B. Garg| A Comparative Study of CPU Scheduling Algorithms

IJGIP
Journal homepage: www.ifrsa.org

A Comparative Study of CPU Scheduling Algorithms

Neetu Goel
Research Scholar,TEERTHANKER MAHAVEER UNIVERSITY
Dr. R.B. Garg
Professor
Delhi School of Professional Studies & Research
(Affiliated to GGSIP University, Delhi)

when there are multiple run-able processes. CPU

ABSTRACT
scheduling is important because it can have a big effect
Developing CPU scheduling algorithms and on resource utilization and the overall performance of
understanding their impact in practice can be difficult the system [2].
and time consuming due to the need to modify and test OS may feature up to 3 distinct types of schedulers: a
operating system kernel code and measure the resulting long term scheduler (also known as an admission
performance on a consistent workload of real scheduler or high level scheduler), a mid-term or
applications. As processor is the important resource, medium-term scheduler and a short-term scheduler (also
CPU scheduling becomes very important in known as a dispatcher or CPU scheduler). The rest of
accomplishing the operating system (OS) design goals. the paper is organized as follows: Section II gives
The intention should be allowed as many as possible review on scheduling algorithms. The proposed
running processes at all time in order to make best use algorithm, Results and Discussion have been given in
of CPU. section II
This paper presents a state diagram that depicts the A. Long-term Scheduler
comparative study of various scheduling algorithms for
The long-term or admission scheduler decides which
a single CPU and shows which algorithm is best for the
jobs or processes are to be admitted to the ready queue;
particular situation. Using this representation, it
that is, when an attempt is made to execute a process its
becomes much easier to understand what is going on
admission to the set of currently executing processes is
inside the system and why a different set of processes is
either authorized or delayed by the long-term scheduler.
a candidate for the allocation of the CPU at different
Thus, this scheduler dictates what processes are to run
time. The objective of the study is to analyze the high
on a system, and the degree of concurrency to be
efficient CPU scheduler on design of the high quality
supported at any one time.
scheduling algorithms which suits the scheduling goals.
Key Words:-Scheduler, State Diagrams, CPU- B. Mid-term Scheduler
Scheduling, Performance The mid-term scheduler temporarily removes process
from main memory and place them on secondary
INTRODUCTION
memory (such as a disk drive) or vice versa. This is
In a single-processor system, only one process can run commonly referred to as “swapping of processes out" or
at a time; any others must wait until the CPU is free and "swapping in" (also incorrectly as "paging out" or
can be rescheduled. The objective of multiprogramming "paging in").
is to have some process running at all times, to C. Short-term Scheduler
maximize CPU utilization [1]. Scheduling is a
The short-term scheduler (also known as the CPU
fundamental operating-system function. Almost all
scheduler) decides which of processes in the ready
computer resources are scheduled before use. The CPU
queue, in memory are to be executed (allocated a CPU)
is, of course, one of the primary computer resources.
next following a clock interrupt, an Input-Output (IO)
Thus, its scheduling is central to operating-system
interrupt and an OS call or another form of signal. Thus
design. CPU scheduling determines which processes run

International Journal of Graphics & Image Processing |Vol 2|issue 4|November 2012 245
 Neetu Goel, R.B. Garg| A Comparative Study of CPU Scheduling Algorithms

the short-term scheduler makes scheduling decisions CPU when it decides to allocate that CPU to another
much more frequent than the long-term or mid-term process, or non pre-emptive (also known as "voluntary"
schedulers. This scheduler can be preemptive, implying or "co-operative"), in that case the scheduler is unable to
that it is capable of forcibly removing processes from a force processes off the CPU.

Swap in Swap out

Partially executed swapped out
processes 4
2
CPU End

Ready Queue

I/O I/O waiting queue 1

3

Figure 1: Shows the following states have been executed in the CPU Scheduler

1. When a process switches from the running which algorithm is judged to be best. The criteria
state to the waiting state. include the following:
2. When a process switches from the running 1. Utilization/Efficiency: keep the CPU busy
state to the ready state. 100% of the time with useful work
3. When a process switches from the waiting state 2. Throughput: maximize the number of jobs
to the ready state. processed per hour.
4. When a process terminates. 3. Turnaround time: from the time of
The success of a CPU scheduler depends on the design submission to the time of completion,
of high quality scheduling algorithm. High-quality CPU minimize the time batch users must wait for
scheduling algorithms rely mainly on criteria such as output
CPU utilization rate, throughput, turnaround time, 4. Waiting time: Sum of times spent in ready
waiting time and response time. Thus, the main impetus queue - Minimize this
of this work is to develop a generalized optimum high 5. Response Time: time from submission till the
quality scheduling algorithm suited for all types of job. first response is produced, minimize response
time for interactive users
SCHEDULING CRITERIA
6. Fairness: make sure each process gets a fair
Different CPU scheduling algorithms have different share of the CPU
properties, and the choice of a particular algorithm may CPU Scheduler whenever the CPU becomes idle; the
favor one class of processes over another. In choosing operating system must select one of the processes in the
which algorithm to use in a particular situation, we must ready queue to be executed. The selection process is
consider the properties of the various algorithms. Many carried out by the short term scheduler (or CPU
criteria have been suggested for comparing CPU scheduler). The scheduler selects from among the
scheduling algorithms. Which characteristics are used processes in the memory that are ready to execute and
for comparison can make a substantial difference in allocates the CPU to one of them Figure 2 shows a
schematic of scheduling:

International Journal of Graphics & Image Processing |Vol 2|issue 4|November 2012 246
 Neetu Goel, R.B. Garg| A Comparative Study of CPU Scheduling Algorithms

Preempted request

Arriving request Scheduled request Server

Scheduler

Pending request

Figure 2: Schematic of Scheduling

The ready queue is not necessarily a first-in, first out case of the processor, the resource to be allocated is
(FIFO) queue. It may be implemented as a FIFO queue, execution time on the processor and the means of
priority queue, a tree or simply an unordered linked list. allocation is scheduling. This way, the scheduler is the
Conceptually, however, all the processes in the ready component of the operating system responsible to grant
queue are lined up waiting for a chance to run on the the right to CPU access to a list of several processes
CPU. (Stalling William, 2004) An operating system ready to execute. This idea is illustrated in the five state
must allocate computer resources among the potentially diagram of figure 3(Galvin et. al. 2001).
competing requirements of multiple processes. In the

New
Scheduler Dispatch
Admit
Ready Running
Interrupt
Release
I/O or Event completion
I/O or Event
wait Exit
Waiting

Figure 3: Life Cycle of Process

In Circumstances first and fourth, there is no choice in

SCHEDULING ALOGRITHMS
terms of scheduling. A new process (if one exists in the
ready queue) must be selected for execution. There is a A. Algorithm and its characteristics
choice, however, in circumstances second and third. The fundamental scheduling algorithms and its
When scheduling takes place under circumstances first characteristics are described in this section.
and fourth, then the scheduling scheme is non- a. First Come First Serve
preemptive; otherwise, the scheduling scheme is The most intuitive and simplest technique is to
preemptive. allow the first process submitted to run first. This
Under non-preemptive scheduling, once the CPU has approach is called as first-come, first-served(FCFS)
been allocated to a process, the process keeps the CPU scheduling. In effect, processes are inserted into the
until it releases the CPU either by terminating or by tail of a queue when they are submitted[1]. The
switching to the waiting state.

International Journal of Graphics & Image Processing |Vol 2|issue 4|November 2012 247
 Neetu Goel, R.B. Garg| A Comparative Study of CPU Scheduling Algorithms

next process is taken from the head of the queue in the four state diagram of figure 4.
when each finishes running. This idea is illustrated

New

Added at First requested process

the tail Running

I/O or Event completion & I/O or Event Release

insert into front of the queue wait

Waiting Exit

Figure 4: First Come First Serve Scheduling

Characteristics The process is allocated to the CPU which has least

 The lack of prioritization does permit every process burst time. A scheduler arranges the processes with the
to eventually complete, hence no starvation. least burst time in head of the queue and longest burst
 Turnaround time, waiting time and response time is time in tail of the queue. This requires advanced
high. knowledge or estimations about the time required for a
 One, Process with longest burst time can process to complete[1]. This algorithm is designed for
monopolize CPU, even if other process burst time maximum throughput in most scenarios. This idea is
is too short. Hence throughput is low [3]. illustrated in the four state diagram of figure 5
b. Non preempted Shortest Job First

New

Added as per First requested process

burst time Running

I/O or Event
I/O or Event completion wait Release

Waiting Exit

Figure 5: Shortest Job First Scheduling

Characteristics penalize processes with high service time requests.
 The real difficulty with the SJF algorithm is, to If the ready list is saturated, then processes with
know the length of the next CPU request. large service times tend to be left in the ready list
 SJF minimizes the average waiting time[3] because while small processes receive service. In extreme
it services small processes before it services large case, when the system has little idle time, processes
ones. While it minimizes average wait time, it may with large service time will never be served. This

International Journal of Graphics & Image Processing |Vol 2|issue 4|November 2012 248
 Neetu Goel, R.B. Garg| A Comparative Study of CPU Scheduling Algorithms

total starvation of large processes is a serious The scheduler goes around this queue, allocating
liability of this algorithm. the CPU to each process for a time interval of
c. Round Robin assigned quantum. New processes are added to the
tail of the queue [4]. This idea is illustrated in the
The Round Robin (RR) scheduling algorithm
four state diagram of figure 6.
assigns a small unit of time, called time slice or
quantum. The ready processes are kept in a queue.

After execution the remaining burst time of

New process > time slice then added to the tail
of ready queue
Added at First requested process
the tail according to the time slice Running

I/O or Event Release

I/O or Event completion
wait
Waiting Exit

Figure 6: Round Robin Scheduling

Characteristics d. Priority Scheduling
 Setting the quantum too short causes too many The O/S assigns a fixed priority rank to each process.
context switches and lower the CPU efficiency. Lower priority processes get interrupted by incoming
 Setting the quantum too long may cause poor higher priority processes. This idea is illustrated in the
response time and approximates FCFS. four state diagram of figure 7
 Because of high waiting times, deadlines are rarely
met in a pure RR system.

If highest priority process will come then added to

New the ready queue as per priority

Highest priority process

Added as per Running
assign priority

Release
I/O or Event completion I/O or Event wait

Waiting Exit

Figure 7: Priority Scheduling

Characteristics Consider the following set of processes, with the length
 Starvation can happen to the low priority of the CPU-burst time in milliseconds is shown in Table
process. 1:
 The waiting time gradually increases for the Process ID Burst Time(ms)
equal priority processes [5]. P0 12
 Higher priority processes have smaller waiting P1 2
time and response time. P2 3
B. Computation of Gantt chart, Waiting Time and P3 2
Turnaround Time P4 6

Table 1: Processes with Its Id and Burst Time

International Journal of Graphics & Image Processing |Vol 2|issue 4|November 2012 249
 Neetu Goel, R.B. Garg| A Comparative Study of CPU Scheduling Algorithms

a. First Come First Serve Turnaround Time (ms)

Process ID
P0 P1 P2 P3 P4
Priority
0 12 14 17 19 25 FCFS SJF Round Scheduling
Robin
Figure 8: Gantt chart for FCFS
P0 12 25 25 20
b. Shortest Job First
P1 14 2 7 2
P1 P3 P2 P4 P0
0 2 4 7 13 25 P2 17 7 10 23

Figure 9: Gantt chart for SJF P3 19 4 12 25

c. Round Robin
Assign time quantum as 5 ms for each process. P4 25 13 23 8

P0 P1 P2 P3 P4 P0 P4 P0
Avg.
0 5 7 10 12 17 22 23 25
Turnaround 17.4 10.2 15.4 15.6
Time
Figure 10: Gantt chart for Round Robin
Table 3: Turnaround Time For Individual Process
d. Priority Scheduling And Average Turnaround Time For Each
Priority is assigned for each process as follows. Scheduling.
Process ID Burst Time(ms) Priority
P0 12 3
Waiting Time (ms)
P1 2 1
Process ID
P2 3 3
Priority
P3 2 4
FCFS SJF Round Scheduling
P4 6 2 Robin
Table 2: Processes with Its Id, Burst Time and
Priority P0 0 13 13 8
P1 P4 P0 P2 P3 P1 12 0 5 0
0 2 8 20 23 25 P2 14 4 7 20
Figure 11 Gantt chart for Priority Scheduling P3 17 2 10 23
For example, turnaround time for the process is P4 19 7 17 2
calculated as time of submission of a process to the time
of completion of the process is obtained through Gantt
chart for SJF scheduling. Turnaround time for process
P0, P1, P2, P3 & P4 is observed as 25,2,7,4 & 13 Avg. 12.4 5.2 10.4 10.6
respectively and average turnaround time is Waiting
(25+2+7+4+13)/5=10.2 ms. Time
The waiting time for the process is calculated as time Table 4: Waiting Time For Individual Process And
taken by the process to wait in the ready queue is Average Waiting Time For Each Scheduling
observed from Gantt chart for SJF scheduling. Waiting
time for process P0, P1, P2, P3 & P4 is obtained as 13, From the above discussion it is clear that First Come
0, 4, 2 & 7 respectively and average waiting time is First Serve (FCFS) & Shortest Job First (SJF) is
(13+0+4+2+7)/5=5.2ms. Similarly the turnaround time generally suitable for batch operating systems and
and waiting time is calculated for all other algorithms Round Robin (RR) & Priority Scheduling (PS) is
and summarized in Table 3 and Table 4. suitable for time sharing systems. No algorithm is
optimum for all type of jobs. Hence it is necessary to

International Journal of Graphics & Image Processing |Vol 2|issue 4|November 2012 250
 Neetu Goel, R.B. Garg| A Comparative Study of CPU Scheduling Algorithms

develop an algorithm with an optimum criteria and [3] Umar Saleem and Muhammad Younus Javed,
suitable for all scenarios. “Simulation of CPU Scheduling Alogrithm”, 0-
7803-6355-8/00/$10.00@2000 IEEE
CONCLUSION & FUTURE WORK
[4] Sun Huajin’, Gao Deyuan, Zhang Shengbing,
The treatment of shortest process in SJF scheduling Wang Danghui; “ Design fast Round Robin
tends to result in increased waiting time for long Scheduler in FPGA”, 0-7803-7547-
processes. And the long process will never get served, 5/021/$17.00@2002 IEEE
though it produces minimum average waiting time and [5] Md. Mamunur Rashid and Md. Nasim Adhtar, “
average turnaround time. It is recommended that any A New Multilevel CPU Scheduling Algorithm”,
kind of simulation for any CPU scheduling algorithm Journals of Applied Sciences 6 (9): 2036-
has limited accuracy. The only way to evaluate a 2039,2009
scheduling algorithm to code it and has to put it in the [6] Sukanya Suranauwarat, “A CPU Scheduling
operating system, only then a proper working capability Algorithm Simulator”, October 10-13, 2007,
of the algorithm can be measured in real time systems. Milwaukee, WI 37th ASEE/IEEE Frontiers in
Education Conference.
REFERENCES [7] Sun Huajin’, Gao Deyuan, Zhang Shengbing,
[1] Silberschatz, A. P.B. Galvin and G. Gagne Wang Danghui;”Design Fast Round Robin
(2012), Operating System Concepts, 8th edition, Scheduler in FPGA”, 0-7803-7547-5/021/$17.00
Wiley India, @ 2002 IEEE.
[2] Sabrian, F., C.D. Nguyen, S. Jha, D. Platt and F. [8] Andrew S. Tanenbaum, Albert S. Woodhull,
Safaei, (2005). Processing resource scheduling in “Operating Systems Design and
programmable networks. Computer Implementation”, Second Edition
communication, 28:676-687 [9] Milan Milenkovic, “Operating System Concepts
and Design”, McGRAW-HILL, Computer
Science Series, Second Edition.

International Journal of Graphics & Image Processing |Vol 2|issue 4|November 2012 251

View publication stats