International Journal on Recent and Innovation Trends in Computing and Communication

Volume: 4 Issue: 4

ISSN: 2321-8169
60 - 64

________________________________________________________________________________________________________

An Improved Min-Min Task Scheduling Algorithm for Load Balancing in Cloud

Computing
Prerit Chawda

Partha Sarathi Chakraborty

M. Tech. (CSE), Dept. of CSE

SRM University, NCR Campus
Ghaziabad, Utter Pradesh, India
[email protected]

Assistant Professor, Dept. of CSE

SRM University, NCR Campus
Ghaziabad, Utter Pradesh, India
[email protected]

Abstract Cloud computing is a model that uses computing resources (CPU, memory, processor etc.) which are delivered as a service over a
network internet. It is a new field for large scale distributed computing. In cloud computing, one of the major task is to execute large number of
tasks with given available resources to achieve high performance, minimal total time for completion, Minimum response time, effective
utilization of resources etc. Keeping all these perspective, we need to design, develop and propose a scheduling algorithm to develop an
allocation map to represent the set of tasks on appropriate resources. Though traditional Min-Min algorithm is a simple, yet efficient algorithm
that provides a better scheduling to minimize the total time for completion of tasks compared to the other algorithms. However the drawback of
this algorithm is the load imbalanced, which is one of the major issue for cloud provider. In this paper, we introduce an improved load balancing
algorithm keeping Min-Min algorithm as base in order to reduce the makespan and increase the resource utilization (ILBMM). The proposed
method has two different phases, in the first phase traditional min-min algorithm is executed and in second phase it selects the task with
maximum completion time and assigns it to appropriate resource to produce better makespan and utilize resource effectively
Keywords-Cloud Computing; Load Balance; Min-Min Algorithm; Task Scheduling

__________________________________________________*****_________________________________________________
I.

INTRODUCTION

Currently Cloud computing has evolved as great potential

technology that is known as a provider of dynamic services
using very large scalable and virtualized resources over the
Internet [5]. Cloud is subject to User Requirement, Load
Balance and other constraints that have direct effect on user
consumption of resources controlled by cloud provider [10].
With increasing number of users on cloud, assuring the
optimum use of cloud resources are difficult task for cloud
service providers and there are many issues associated with the
task scheduling algorithm and so the load balancing can be
used for cloud for balancing the load.
Load balancing [1] is a method that distributes the workload
among diverse nodes in the given environment such that it
ensures no node in the system is over loaded or sits idle for
any instant of time. An efficient load balancing algorithm will
make sure that every node in the system does more or less
same volume of work.
The responsibility of load balancing algorithm is that to map
the jobs which are set forth to the cloud domain to the
unoccupied resources so that the overall available response
time is improved as well as it provides efficient resource
utilization. Balancing the load became one of the crucial
concerns in cloud computing since we cannot predict the
number of requests that are issued at each second in cloud
environment. The unpredictability is due to the ever changing
behavior of the cloud. The main focus of load balancing in the
cloud domain is in allocating the load dynamically among the
nodes in order to satisfy the user requirements and to provide
maximum resource utilization by assorting the overall available
load to distinct node

Fig 1 Diagram for load balancing

A. Task scheduling in cloud computing
In cloud, many tasks have to be executed by the available resources
to achieve distinct intensions. There is need for a scheduling
algorithm that is used by task scheduler to outperform appropriate
allocation map of task on resources [11].
There are different types of scheduling are based on different
criterias, such as Static and Dynamic scheduling, Centralize and
Distributed scheduling, Preemptive and Non-Preemptive scheduling,
Cooperative scheduling, Immediate and Batch mode scheduling
[12][14][15].
Cloud task scheduling is an NP-complete problem in general [7]. In
the typical cloud scenario, cloud users submit their tasks to cloud
scheduler. The scheduler on cloud will first query the Cloud
Information Service to gather the information regarding the
availability and properties of resources, and then they schedule the
tasks on the resources which satisfy the tasks requirements. Once task
is executed, the result is sent to the users.
Scheduling of tasks in such cloud environment is the challenging task
because of its high heterogeneity in operating systems, architecture,
cloud providers and resource consumers [3].
The main purpose of a cloud task scheduling algorithm is to shorten
overall the completion time of all tasks submitted by users, enhance
the utilization of cloud resources and satisfy requirements of different
users [2].In this paper a new task scheduling algorithm is proposed to

60
IJRITCC | April 2016, Available @ http://www.ijritcc.org
____________________________________________________________________________________________________________________

International Journal on Recent and Innovation Trends in Computing and Communication

Volume: 4 Issue: 4

ISSN: 2321-8169
60 - 64

________________________________________________________________________________________________________
decrease jobs completion time and improve the load balance in
in the system and n is the number of submitted tasks which
the cloud
should be scheduled [4], [6]
II.

RELATED WORK

A load balance scheduling algorithm aims to increase the utilization

of resources with light load or idle resources thereby freeing the
resources with heavy load. Main goal of this algorithm is to distribute
load among all available resources and also to minimize the
makespan which effectively utilizes the resources.
In the literature, large numbers of task scheduling algorithm were
proposed in the past. Braun et al [7] have studied the relative
performance of eleven heuristic algorithms for task scheduling such
as Opportunistic Load Balancing (OLB), Minimum Execution Time
(MET), Minimum Completion Time (MCT), Min-Min, Max-Min,
Duplex, Genetic Algorithm (GA), etc. They have also provided a
simulation basis for researchers to test the algorithms. Their results
show that the simple Min-Min algorithm produces a better schedule
that minimizes the makespan than the other algorithms and performs
next to GA which the rate of improvement is also very small in most
of the scenarios. In this section different task scheduling techniques
are reviewed briefly
OLB allocates task to the resource that are to be ready next, without
any execution time of task consideration on that resource [8]. If more
than one resource becomes ready than, the resources are arbitrarily
chosen. The intention behind OLB is to keep all machines as busy as
possible one advantage of OLB is its simplicity, but because OLB
does not consider task execution times, so it will generate poor
makespan.

III.

TRADITIONAL MIN-MIN SCHEDULING ALGORITHM

The Min-Min algorithm is simple and still basis of present cloud

scheduling algorithm [13]. It starts with a set U of all unmapped
tasks. Then the resource R which has the minimum completion time
for all tasks is found. Next, the task T with the minimum size is
selected and assigned to the corresponding resource R (hence the
name Min-Min). Last, the task T is removed from set U and the same
procedure is repeated by Min-Min until all tasks are assigned (i.e., set
U is empty).
The algorithm works better when the situation is like where the small
tasks are greater in number of than the large tasks. The algorithm has
a disadvantage that it leads to starvation. .Min-Min is a simple and
fast algorithm capable of providing improved performance. Min-Min
schedules the ideal tasks at first which results in best schedules and
improve the overall make span. Assigning small task first is its
drawback. Thus, smaller tasks will get executed first, while the larger
tasks keeps on in the waiting stage ,which will finally results in poor
machine use.
The pseudo code of Min-Min algorithm is represented in Fig 2
assuming we have a set of n tasks (T1, T2, T3 Tn) need to be
scheduled onto m available resources (R1, R2, R3 Rm). We
denotes the Expected Completion Time for task i on resources j as
Ctij that is calculated as Ctij = Etij + rtj, Where rtj represents the
Ready Time of resource Rj and Etij represents the Execution Time of
task Ti on resource Rj.

MET assigns each task to the resource that performs it in the least
amount of execution time, no matter whether this resource is
available or not at thattime[10]. This heuristic can cause a severe load
imbalance across the resources. However, this is one of the heuristics
that is implemented in SmartNet. The aim behind MET is to give
each task to its best machine
MCT assigns each task to the resource which obtains earliest
completion time for that task, which causes some of the tasks to be
assigned to the resources whose execution time is not minimum for
them [9]. The intension of MCT is to combine the benefits of OLB
and MET while avoiding the drawbacks of these two algorithms.
Min-Min algorithm starts with the set of all the unmapped tasks.
Machine with the least or minimum completion time for all jobs will
be selected. Next the job whose overall completion time is minimum
will be selected and will be mapped to the respective resource. Ready
time of the resource will then be updated. Repeat this process until all
the unmapped tasks are assigned. This algorithm considers all jobs at
any time compared to the MCT, which shows that it produces a better
makespan.
Max-Min is similar to Min-Min algorithm. The machine that has the
minimum completion time for all jobs is selected. Then the job with
the overall maximum completion time is selected and will be mapped
to the resource. Next the ready time of the resource is updated.
Repeat this process until all the unmapped tasks are alocated. The
basic idea of this algorithm is to reduce the overall waiting time of
the large jobs
SaeedParsa et al.proposed a new task scheduling algorithm called
RASA [4]. It takes advantage of both Max-min and Min-min
algorithm. RASA uses the Min-min strategy to execute small tasks
before large ones and applies the Max-min strategy to avoid delays in
the execution of the large tasks and to support concurrency in the
execution of large and small tasks
Each of Max-min and Min-min algorithms have running time
complexity of (mn2), where m is the number of resources currently

Fig 2 Flow chart of Min-Min algorithm.

61

IJRITCC | April 2016, Available @ http://www.ijritcc.org

____________________________________________________________________________________________________________________

International Journal on Recent and Innovation Trends in Computing and Communication

Volume: 4 Issue: 4

ISSN: 2321-8169
60 - 64

________________________________________________________________________________________________________
A. Algorithm
//Rescheduling to balance the load sort resources in order
of completion time
7. Do while the most heavy load resource is considered no
//Calculate the completion time matrix for all task on each
need for rescheduling
resource
8.
Find task Ti that cost minimum execution time on the
1. For all submitted tasks in task set; T i
heavy load resource Rj
1.1. For all resources; Rj
9.
Find the minimum completion time of T i produced by
1.1.1. Ctij = Etij + rtj // Ctij means expected
resource Rk
completion time of task i on resource j
10. If such minimum completion time <makespan
2. Do while tasks set is not empty
11. Reassign Task Ti to Resource Rk
3. Find task Tkthat cost minimum execution time
12. Update the ready time of both Rj and Rk
4. Assign task Tk to resource Rj which gives minimum
13. End If
Expected completion time
14. End Do
5. Remove task Tk from tasks set.
//where Makespan represents maximum completion time
5.1. Update rtj for selected Rj.
of all tasks which equals to the completion time of the
5.2. Update Cij for all Ti
most heavy load resource
6. End do
IV.

IMPROVED LOAD BALANCE MIN-MIN ALGORITHM

In this paper, the Load Balance Improved Min-Min (ILBMM)

scheduling algorithm is proposed that takes the characteristic
of the Min-Min scheduling algorithm as foundation. The
performance of Min-Min scheduling algorithm is considered
to minimize the completion time of all works. However, the
biggest weakness of Min-Min algorithm is it does not
considers the work load of each resource. Therefore, some
resources maybe always get busy but some nodes maybe still,
as shown in Figure 2. The proposed ILBMM will improve the
load unbalance of the Min-Min and reduce the execution time
of each resource effectively
The pseudo code of ILBMM algorithm is represented in Fig 3.
It starts by executing Min-Min algorithm. After execution of
the concerned algorithm, it chooses the smallest size task from
the most heavy load resource and calculates the completion
time for that task on all other resources. Afterwards, the
minimum completion time of the task is compared with the
makespan produced by the algorithm. If it is less than
makespan then the task is reassigned to the resource that
produce it, and the ready time of both resources are updated.
The process repeats until no other resources can produce less
completion time for the smallest task on the heavy load
resource than the makespan. Thus the heavy load resources are
freed and the light load or idle resources are more utilized.
This makes ILBMM to produce a schedule which improves
load balancing and also reduces the overall completion time.
A. Algorithm
//Calculate the completion time matrix for all task on each
resource
1. For all submitted tasks in the set; Ti
1.1. For all resources; Rj
1.1.1. Ctij=Etij+rtj; //Cij means expected completion
time of task i on resource j
2. Do while tasks set is not empty
3. Find task Tk that cost minimum execution time.
4. Assign Tk to the resource Rj which gives minimum
expected complete time
5. Remove Tk from the tasks set
5.1. Update ready time rtj for select Rj
5.2. Update Cij for all Ti
6. End Do

Fig 3 Flow Chart of ILBMM

62
IJRITCC | April 2016, Available @ http://www.ijritcc.org
____________________________________________________________________________________________________________________

International Journal on Recent and Innovation Trends in Computing and Communication

Volume: 4 Issue: 4

ISSN: 2321-8169
60 - 64

________________________________________________________________________________________________________
V. THEORETICAL ANALYSIS
Assume that task scheduler has meta-tasks and resource as
given below. Table 1 represents the volume of instruction and
data of task T1to T4. Instruction volume is specified in MI
(Millions instructions) Unit and data volume is specified in
MB.
Table 1 Task specification
Task

Instruction
volume (MI)

Data volume
(MB)

T1

8178

137

T2

11295

258

T3

12109

182

T4

6107

137

Table 2 represents processing speed and bandwidth of

communication links for all resources where processing speed
is specified in MIPS and bandwidth is specified in Mbps.

Figure above represent that the ILBMM gives better load

balance and makespan than Min-Min algorithm

Table 2 Resource specification

Resource

Processing
speed (MIPS)

Bandwidth
(Mbps)

R1

100

70

R2

350

60

Using Data given in table 1 and 2 to calculate the expected

execution time of the task on each of the resource
Table 3 Execution time of tasks on each resource.
Task

Resources
R1

Fig 4 Makespan (Total completion time of all task)

R2

T1

81.78

23.36

T2

112.95

32.27

T3

121.09

34.59

T4

61.07

17.45

Table 3 demonstrates calculated execution time of various task

on various resources data in table 3 will be updated after all
tasks are allocated.
In above scenario, Improved Min-Min achieves total
makespan equal to 84.31 seconds and Min-Min achieves
makespan equals 90.22 seconds

VI. CONCLUSION AND FUTURE WORK

To achieve high computing throughput in a cloud
environment, anew scheduling algorithms, ILBMM were
proposed in this paper. The Theoretical analysis show that
under all possible situations the ILBMM is capable of
decreasing completion time of tasks, improving load balance
of resources and gain better overall performance than MinMin algorithm.
This paper is only concerned with the makespan and load
balancing for task scheduling based on Min-Min algorithm in
Cloud environment. In future for simulation, we will use
CloudSim which is java based simulation toolkit that enables
modelling, simulation and experimenting on designing cloud
computing infrastructures
REFERENCES
[1] N. Ajith Singh, M. Hemalatha, An approach on semi distributed
load balancing algorithm for cloud computing systems
International Journal of Computer Applications Vol-56 No.12
2012
[2] RS Chang, CY Lin, and CF Lin. An Adaptive Scoring Job
Scheduling algorithm for grid computing, Information Sciences
Elsevier, Volume 207, Pages 7989, 2012
[3] T Kokilavani, GA DI. Load Balanced Min-Min Algorithm for
Static Meta-Task Scheduling in Grid Computing, International
Journal of Computer Applications, Number 2 - Article 7, 2011
[4] SaeedParsa and Reza Entezari-Maleki , "RASA: A New Grid
Task Scheduling Algorithm", International Journal of Digital
Content Technology and its Applications,Vol. 3, pp. 91-99, 2009
[5] SalimBitam, "Bees life algorithms for job scheduling in cloud
computing", International Conference on Computing and
Information Technology, 2012
[6] R. F. Freund, M. Gherrity, S. Ambrosius, M. Campbell, M.
Halderman, D. Hensgen, E. Keith, T. Kidd, M. Kussow, J. D.
Lima, F. Mirabile, L. Moore, B. Rust and H. J. Siegel,
"Scheduling Resource in Multi-User, Heterogeneous, Computing
Environment with SmartNet,"In the Proceeding of the Seventh
Heterogeneous Computing Workshop, 1998

63
IJRITCC | April 2016, Available @ http://www.ijritcc.org
____________________________________________________________________________________________________________________

International Journal on Recent and Innovation Trends in Computing and Communication

Volume: 4 Issue: 4

ISSN: 2321-8169
60 - 64

________________________________________________________________________________________________________
[7] Braun, Tracy D, et al. "A comparison of eleven static heuristics
for mapping a class of independent tasks onto heterogeneous
distributed computing systems" Journal of Parallel and
Distributed computing , Volume 61, Issue 6, Pages 810 837,
2001
[8] AnjuBala, Dr.InderverChana, "A Survey of Various Workflow
Scheduling Algorithm in Cloud Environment",2nd national
conference
on
information
and
communication
technology(NCICT)2011
[9] George Amalarethinam. D.I, VaaheedhaKfatheen .S, Max-min
Average Algorithm for Scheduling Tasks in Grid Computing
Systems, International Journal of Computer Science and
Information Technologies, Vol. 3 (2) , 2012 ,3659-3663
[10] Elzeki, O. M., M. Z. Reshad, and M. A. Elsoud. "Improved MaxMin Algorithm in Cloud Computing", International Journal of
Computer Applications, Volume 50, Issue 12, Pages 22-27, 2012
[11] S.-C.Wang, K.-Q. Yan, S.-S.Wang, C.-W. Chen, A three-phases
scheduling in a hierarchical cloud computing network, in:

[12]

[13]

[14]
[15]

Communications and Mobile Computing (CMC), 2011 Third

International Conference on,IEEE, 2011, pp. 114117
D.I. George Amalarethinam and P. Muthulakshmi, "An Overview
of the scheduling policies and algorithms in Grid Computing ",
International Journal of Research and Reviews in Computer
Science, Vol. 2, No. 2, pp. 280294, 2011.
Yu, Xiaogao, and Xiaopeng Yu. "A new grid computation-based
Min-Min algorithm", Fuzzy Systems and Knowledge Discovery,
FSKD'09 Sixth International Conference on, Volume 1, Pages 43
45, IEEE, 2009
FatosXhafa, Ajith Abraham, "Computational models and
heuristics methods for grid scheduling problems", Future
Generation Computer systems, Vol. 26, pp. 608-621, 2010.
T. Casavant and J. Kuhl, "A Taxonomy of scheduling in General
purpose distributed computing systems", IEEE Trans on Software
Engineering, Vol. 14, No. 2, pp. 141154, 198.

64
IJRITCC | April 2016, Available @ http://www.ijritcc.org
____________________________________________________________________________________________________________________