International Journal of Technical Research and Applications e-ISSN: 2320-8163,

www.ijtra.com Volume-2, Special Issue 2 (July-Aug 2014), PP. 01-04

1 | P a g e
MJHP - JOB SCHEDULING ALGORITHM FOR
CLOUD ENVIRONMENT
S.Rekha, R.Santhosh Kumar
Department of Information Technology
Sri Venkateswara College of Engineering
Chennai, India

Abstract- Cloud computing can be defined as a type of Internet-
based computing, where different services such as servers,
storage and applications are delivered to an organization's
computers and devices through the Internet. It is a cluster of
network where the nodes interact to accomplish a big
computational task. Due to its efficiency in running multiple
programs simultaneously, cloud computing is emerging as a
popular domain in technology. Intense research has thrown
light on how the resources are shared and jobs are scheduled
amongst the nodes. A proper job-scheduling algorithm is
required for the efficient functioning of the cloud environment.
The proposed priority based scheduling algorithm for cloud
computing is based on factors that govern the functioning of a
job.
Keywords- Cloud Computing, Job Scheduling, Priority,
Computational Complexity and Level of Parallelism.
I. Introduction
With the rapid evolving technology, more business
organizations are adapting to cloud computing[3] because of
its merits such as the ability to process large amount of data
and perform complex computations. Thus for a cloud
environment to work effectively, proper scheduling has to be
designed and implemented as job delays or data loss can
prove to be a major hazard for the organization. A cost
effective and maximum performance achievable algorithm
will prove to be the backbone of the cloud environment. The
existing Batch mode heuristic scheduling algorithms
(BMHA) are: First Come First Served scheduling algorithm
(FCFS), Shortest Job Fastest Resource (SJFR), Longest Job
Fastest Resource (LJFR), MinMin algorithm and MaxMin
algorithm. These algorithms consider all jobs with equal
importance. The existing Priority Based scheduling
Algorithm[1] gives first preference to a job that has highest
computational complexity and level of parallelism. However,
this algorithm faces the drawback of time wasted in waiting
for the high priority job to be executed completely. This time
delay can be avoided if medium[2] level job is executed prior
to the high priority job. This is the main focus of the
proposed MJHP - Medium Job High Priority job scheduling
algorithm[7]. Thus, an ideal scheduling[6] algorithm can be
developed that satisfies the time constraint as well as
achieves best performance. The paper is organized as
follows: section 2 deals with the cloud architecture, section 3
describes the proposed algorithm, section 4 provides a picture
of performance analysis and section 5 concludes the paper.

II. Cloud Architecture
The cloud architecture comprises of the cloud components
that communicative with each other and delivers the output.
The frontend platforms visible to the user can be a mobile
device or web and back end platform components are servers,
storage and a network. This is shown in fig 1.

Fig 1. Cloud Computing Architecture
Jobs arrive at the scheduler to be executed along with request
for cloud resources[3]. The function of the scheduler is to
select how several incoming jobs have to be processed and
allocate resources wisely. The overall performance and
throughput of the system depends on how the scheduler
works. The scenario is depicted in fig 2.


Fig 2 : Cloud job scheduler.
III. Proposed algorithm - MJHP Scheduling in Cloud
Computing
Jobs arriving at the scheduler are classified as high, medium
and low based on the computational complexity of the job
and level of parallelism of the resources. The level of
parallelism of a resource and computational power of a job is
decided by considering the job parallelism, resource
parallelism and jobs computational complexity respectively.
In the exisiting algorithm, higher priority was assigned to job
of higher computational complexity and the resource
exhibiting higher level of parallelism .The fastest resource
available was assigned to the job of high priority. This
method of allocation of resources to the job with high
priority[9] resulted in significant waste of time as the
processor has to wait till the complex job has to be executed.
This threat is overcome, in the proposed MJHP algorithm
where preference is given to a job with medium
International Journal of Technical Research and Applications e-ISSN: 2320-8163,
www.ijtra.com Volume-2, Special Issue 2 (July-Aug 2014), PP. 01-04
2 | P a g e
[2]computational complexity. This priority algorithm
optimizes the computational speed of the cloud[4] and
reduces the usage of nodes and also shows a consistent
performance during execution of the assigned jobs.
A. Computational Complexity
The Directed Acyclic Graph (DAG) gives a visual
representation of the task partitioning [10]algorithm which
efficiently divides a job into subtasks of appropriate grain
size. This graph is shown in fig 3. Each node in DAG
represents sequence of operations. Each task[10] can be
executed on a processor and the directed arc depicts the
transfer of relevant data from one processor to another. The
amount of computations involved in a particular node is
represented by node weight.

Fig 3 : Directed Acyclic Graph (DAG)

This graph needs to be traversed to find out the longest path.
The total sum of the amount of computations involved in
each node through which the traversal has been performed
leads to computational complexity of the application.
B. Level Of Parallelism
Generally, the amount of parallelism exhibited by a job is
computed and analyzed by analyzing its layered DAG
representation. The width of the DAG is equal to the number
of sub tasks, which are executed through parallelism. The
maximum number of independent instructions getting
executed in a unit time (in one clock cycle) is equal to the
width of the DAG that gives the amount of parallelism
exhibited by the job. The amount of parallelism exhibited by
a resource is computed by considering the number of
operations per cycle per processor, number of processors per
node and number of nodes in a system. The amount of
parallelism exhibited by each free resource available in the
cloud is computed. The amounts of parallelism exhibited by
all the available free resources in the cloud are fixed by
analyzing the max, min and mid ranges. The value for the
level of parallelism is assigned by comparing the amount of
parallelism exhibited by each job with the max, min and mid
ranges.
C. Priority Assignment
In the proposed MJHP algorithm, higher priority is given to a
job which has high computational complexity and level of
parallelism because time can be saved if the medium level
jobs[2] are executed first. Medium priority is assigned
generally to a job which needs high computational power and
which exhibits high parallelism. A job, which exhibits low
parallelism and needs low computational power for execution
is given a low priority. The fastest free resource available in
the cloud is allocated to the job which has higher priority.
The procedure is given below :

Amount of Parallelism = OC* PN*NS
Where OC= No. of operations per cycle per processor
PN= No.of processors per node
NS=No. of nodes in a cloud.
Let m represent number of free resources available in the
cloud and n represent the number of jobs present in the
queue. The worst case time complexity of the algorithm is
O(n logn) ,when m <= n and O(m logm) when m > n.

D. Proposed MJHP Algorithm
Step 1 : AssignLevelofParallelism( ResourceList Rs_List)
While(Rs_List!=NULL)
For each resource
/*OC = No. of operations per cycle per processor
PN = No. of processors per node
NS = No. of nodes in a coud*/
/* LL_List contains the amount of parallelism
Exhibited by each resource */
LL_List[i] = OC*PN*NS
End While
Find the Max, Min and Mid values in PR_List
/* LJ_List contains the amount of parallelism
exhibited
by each job */
For each job in LJ_List
If LJ_List[i] >= Maximum
LP_List[i] = High //LP_List contains the level of
parallelism value
Else If LJ_List[i] >= Middle
LP_List[i] = Medium
Else LP_List[i] = Low
EndIf
End AssignLevelofParallelism
Step 2 :Assign Priority Procedure
AssignPriority ( CloudList CL_List)
While( CL_List !=NULL)
For each job
/* CompC_List contains the Computational
Complexity of jobs */
If (CompC_List[i] =Medium AND LP_List[i] =
Medium)
Priority[i] = 1
Else If (CompC_List[i] = Medium AND LP_List[i]
= High)
Priority[i] = 2
Else If (CompC_List[i] = Medium AND LP_List[i]
=Low)
Priority[i] = 3
Else If (CompC_List[i] = High AND LP_List[i] =
Medium)
Priority[i] = 4
Else If (CompC_List[i] = High AND LP_List[i] =
High)
Priority[i] = 5
Else If (CompC_List[i] = High AND LP_List[i] =
Low)
Priority[i] = 6
International Journal of Technical Research and Applications e-ISSN: 2320-8163,
www.ijtra.com Volume-2, Special Issue 2 (July-Aug 2014), PP. 01-04
3 | P a g e
Else If (CompC_List[i] = Low AND LP_List[i] =
Medium)
Priority[i] = 7
Else If (CompC_List[i] = Low AND LP_List[i] =
High)
Priority[i] = 8
ElseIf (CompC_List[i] = Low AND LP_List[i] =
Low)
Priority[i] = 9
EndIf
End AssignPriority
IV. PERFORMANCE STUDY
In this section, a performance study is carried out between
the algorithms mentioned so far - First Come First Served
scheduling algorithm (FCFS),Shortest Job Fastest Resource
(SJFR), Longest Job Fastest Resource (LJFR), MinMin
algorithm, MaxMin algorithm, Priority Based Scheduling
[3]and MJHP Job Scheduling algorithm.

A. First Come First Served (FCFS)
FCFS is a very basic job scheduling [7]algorithm which
allocates resources to jobs as they arrive. It does not consider
factors like computational complexity or level of parallelism.
Hence its performance is very low. This is shown in Fig 4.


Fig 4:Implementation of FCFS

B. Shortest Job Fastest Resource (SJFR)
Shortest Job Fastest Resource is a scheduling algorithm,
assigns the job with very low turnaround time to the fastest
resources in the cloud. From the Fig 5 we can decipher that
SJFR is more stable in handling jobs and hence outperforms
FCFS scheduling algorithm.

Fig 5. Implementation of SJFR

C. Longest Job Fastest Resource (LJFR)
Longest Job Fastest Resource is a scheduling algorithm that
assigns the complex job to a big efficiency resource. It tries
to reduce the overall execution time of the jobs. From the Fig
6 of the LJFR algorithm we can infer that LJFR outperforms
FCFS and the SJFR as the jobs of high computational
complexity are assigned to faster resources in the cloud
which leads to shorter execution time.

Fig 6 - Implementation of LJFR

D. Min-Min Algorithm
The Min-Min algorithm schedules the less complex jobs to
high performance resources for execution. From the fig 7 we
can observe that outperforms FCFS and RR but shows low
performance comparing the other algorithms due to the delay
caused in execution of complex jobs.

Fig 7- Implementation of Min-Min

E. Max- Min Algorithm
The complex job is scheduled first to high performance
resources in the cloud and leads to the long delay in the
execution of less complex jobs. Fig 8 shows the performance
of the Max-Min algorithm where it outperforms FCFS, RR,
MIN-MIN algorithms.

Fig 8- Implementation of Max- Min

F. Priority Based Scheduling
The priority based scheduling [3]gives high priority to a job
with maximum computational complexity and level of
parallelism. Even though it outperforms FCFS, SJFR, LJFR,
Min-Min and Max-Min it still has a relatively low
performance than MJHP job scheduling algorithm[7] as some
amount of time is wasted when complex jobs are executed
first.
International Journal of Technical Research and Applications e-ISSN: 2320-8163,
www.ijtra.com Volume-2, Special Issue 2 (July-Aug 2014), PP. 01-04
4 | P a g e

Fig 9- Implementation of Priority Based Scheduling

G.MJHP Job Scheduling
The MJHP job scheduling algorithm gives high priority to a
job with medium computational complexity and medium
level of parallelism. It outperforms the existing FCFS, SJFR,
LJFR, Min-Min and Max-Min and the Priority Based
Scheduling algorithms[3]. The MJHP job scheduling
algorithm assigns the medium priority jobs with medium
computational complexity and medium level of parallelism to
the fastest resource . Hence the waiting time of jobs with
smaller computational complexity is avoided.

Fig 10- Implementation of MJHP Job Scheduling

H. Comparative Study Between MJHP Job Scheduling
Algorithm And Other Algorithms
From the above graphs, it is evident that MJHP has a higher
efficiency than other algorithms. It leads to much better
utilization of [9]. Factors that govern the performance of the
system such as computational complexity and level of
parallelism of a resource are properly analyzed and jobs are
classified as high, medium and low. First preference is given
to medium level jobs as the time taken to be completed is less
than high level jobs. Therefore, overall performance of the
system is improved. It outperforms existing algorithms like
FCFS, SJFR, LJFR, Min- Min, Max- Min and Priority Based
job scheduling[3].

Fig 11- Comparative Study graph

From fig 11, it is notable that for an ideal cloud
environment[3], MJHP scheduling proves to be the best as it
provides maximum achievable throughput in a minimum
amount of time making the network to work faster and more
reliable.
V. Conclusion
Cloud computing[3] has evolved as a dynamic technology
where more and more organizations are relying on it for
various services such as email, file sharing, customer
relationship management, storage and so on. Resources [9]
such as servers, storage, network, applications and processes
can be dynamically shaped or craved out from the underlying
hardware infrastructure and made available to a workload. A
proper scheduling algorithm is required for accomplishing
the goals of the services. The proposed MJHP job scheduling
algorithm satisfies the necessary constraints and proves to be
very efficient than other existing algorithms. The reliability
and consistency of the proposed MJHP algorithm is proved
through simulation results and its superiority over other
known algorithms is depicted.

REFERENCES
[1] S. Rekha and Santhosh Kumar. R, Priority Based Job
Scheduling for Heterogenous Cloud Environment, IJCSI
International Journal Of Computer Science, Volume 11, Issue 3
May 2014.
[2] Dr. G. Sumathi, R. Santhosh Kumar, and S. Sathyanarayanan,
MidSFN Local Scheduling Algorithm for Heterogeneous Grid
Environment, IJCSI International Journal of Computer Science
Issues, Vol. 9, Issue 3, No 3, May 2012
[3] Shamsollah Ghanbari, Mohamed Othman, A Priority based Job
Scheduling Algorithm in Cloud Computing, Procedia
Engineering 50 (2012) 778 785.
[4] Isam Azawi Mohialdeen, COMPARATIVE STUDY OF
SCHEDULING ALGORITHMS IN CLOUD COMPUTING
ENVIRONMENT, Journal of Computer Science, 9 (2): 252-
263, 2013, ISSN 1549-3636
[5] Stelios Sotiriadis, Nik Bessis, Nick Antonopoulos, Towards
inter-cloud schedulers:A survey of meta-scheduling approaches
,2011 International Conference on P2P, Parallel, Grid, Cloud
and Internet Computing
[6] Mladen A. Vouk, Cloud Computing Issues, Research and
Implementations,Journal of Computing and Information
Technology - CIT 16,2008,4,235246doi:10.2498/cit.1001391
[7] Yun-Han Lee et al, Improving Job Scheduling Algorithms in a
Grid Environment, Future Generation Computer Systems,
27(2011) 991998
[8] Wei Wang, Cloud-DLS: Dynamic Trusted Scheduling for Cloud
Computing, Expert Systems with Applications 39 (2012) 2321
2329.
[9] Tai-Lung Chen et al, Scheduling of Job Combination and
Dispatching Strategy for Grid and Cloud System, GPC,(2010)
612621.
[10] Monir Abdullah, Mohamed Othman et al, Optimal
Workload Allocation Model for Scheduling Divisible Data Grid
Applications,Future Generation Computer Systems 26 (2010)
971-978.