A novel load balancing model for overloaded cloud

IJRET: International Journal of Research in Engineering and Technology eISSN: 2319-1163 | pISSN: 2321-7308
__________________________________________________________________________________________
Volume: 03 Special Issue: 07 | May-2014, Available @ http://www.ijret.org 412
A NOVEL LOAD BALANCING MODEL FOR OVERLOADED CLOUD
PARTITION
Mithra P B1
, P Mohamed Shameem2
1
Mtech Student, Dept of CSE, TKM Institute of Technology, Kerala, India
2
Associate Professor, Dept of CSE, TKM Institute of Technology, Kerala, India
Abstract
Load balancing is an efficient solution that distributes excess workload evenly to all nodes in cloud environment. The Load
balancing model is used for the public cloud having numerous nodes in different geographic locations. The model divides public
cloud into several cloud partitions. When partition status becomes overloaded, cloud partitioning is repeated. It reduces the working
efficiency and expected response time of the system. To overcome this issue we propose a novel load balancing strategy for
overloaded cloud partition. The overloaded load balancing strategy maintains two queues at overloaded condition. Priority queue is
used when the cloud partition status is idle and normal and non priority queue is used when partition status is overloaded. At
overloaded condition an overloaded partition scheduling algorithm is used for the allocation of jobs to Non priority queue. When a
priory job ends, then one of the jobs in Non priority queue moves to priority queue based on arrival time and processing power
required
Keywords: Cloud Partition, Job Splitting Index, Non priority queue, Overloaded partition, Priority queue
----------------------------------------------------------------------***------------------------------------------------------------------------
1. INTRODUCTION
Cloud computing is an emerging technology that brings many
changes to the IT industry. Cloud computing allow users to
take advantage from all these technologies, without deep
knowledge about or expertise with them. Load balancing
schemes depending on whether the system dynamics are
important can be either static or dynamic [1]. It is an
efficient solution that distributes excess workload evenly to
all nodes in cloud environment [2]. The load balancing model
is used for the public cloud having numerous nodes in
different geographic locations [3]. The model for such a
cloud computing environment leads to high cost when there is
an increase in number of nodes. It is also difficult for the
existing load balancing strategies to apply when the
environment is large and complex. So cloud partitioning is
chosen that divides the public cloud into several cloud
partitions by the random selection of nodes. The model
includes main controller and partition balancers to perform
load balancing solution. When the cloud partition status is
overloaded, cloud partitioning is repeated. It reduces the
working efficiency and expected response time of the system.
In the proposed load balancing strategy each node maintains
two queues Priority and Non-priority queue. It is a modified
approach of existing load balancing model. Priority queue is
used when the cloud partition status is idle or normal and Non
priority queue is used when partition status is overloaded. At
overloaded condition the jobs in idle and normal partition
status are moved to Priority queue and the jobs after
overloaded status are moved to non-priority queue. For the
better allocation of jobs at overloaded situation we propose an
overloaded partition scheduling algorithm. The main features
of our algorithm can be listed as follows:
 Minimum response time at overloaded situation
 Provides better fault tolerance
 Simplifies load balancing
The rest of the paper is organized as follows: In section II, we
survey related works of load balancing in cloud computing
environment. In section III we do the proposed work. In
section IV we do the performance analysis on our proposed
algorithm. Finally, in section V summarizes our findings and
concludes the paper.
2. RELATED WORK
Cloud computing has attracted considerable research attention,
but only a small portion of the work has been done so far.
There has also been much research in towards different styles
of load balancing. Here, we survey those that proposed certain
techniques and algorithms for load balancing in cloud
environment.
Martin Randles, David Lamb (2010)[4] investigates three
viable methods for load balancing. Firstly, nature-inspired
algorithms for achieving global load balancing. Secondly, load
balancing of all system nodes using random sampling of the
system domain. Thirdly, optimizes job assignment by
connecting similar services by local re-wiring.

__________________________________________________________________________________________
Kumar Nishant (2012)[5] proposed an algorithm for effective
distribution of workloads among the nodes of a cloud
environment by the use of Ant Colony Optimization (ACO).
This is a modified approach of ant colony optimization. The
ACO is used for load balancing. The main advantage of this
approach is the detection of overloaded and under loaded
nodes. Nidhi Jain Kansal (2012)[6] study the existing load
balancing techniques in cloud computing and further compares
them based on various parameters like performance, scalability,
associated overhead etc that are considered in different
techniques.
Shantanu Dutt (1993)[7] presents a very efficient graph
partitioning scheme that uses the basic strategy of the
Kernighan-Lin algorithm to swap pairs of nodes to improve an
existing partition of a graph G. The algorithm attempts to find a
partition of a set of nodes (V) into disjoint subset A, B of equal
sizes such that sum of the weights of the edges between nodes
in A and B is minimized. For that take the initial partition and
iteratively improve it. Vertex pairs with largest decrease or
smallest increase in cut size are exchanged. These vertices are
then locked. This process continues until all vertices are
locked.
Tarun Kumar (2012)[8] proposed Load Balanced Max Min
algorithm. The proposed algorithm outperforms Max-Min
because it focuses on minimizing the completion time of tasks.
The proposed algorithm is executed in two-phases. It uses the
advantages of Max- Min and covers its disadvantages by
reducing makespan and maximizing resource utilization.
Gaochao Xu (2013)[1] proposed a better load balancing model
for public cloud based on the cloud partitioning. The model
includes Main Controller and Balancers to perform load
balancing solution. The Main Controller selects the best cloud
partition and Balancers choose right load balancing strategy to
distribute the jobs to cloud partition. Here, the idle partition
status uses an improved Round Robin algorithm and the
normal status uses a Game theory based load balancing
strategy. When partition status becomes overloaded, cloud
partitioning is repeated. It reduces the working efficiency and
expected response time of the system
In reference to [1], we modified the load balancing model by
maintaining two queues at overloaded condition and use a
scheduling algorithm for the allocation of jobs in the way
mentioned below.
3. PROPOSED WORK
The load balancing model is used for the public cloud which
has numerous nodes in many different geographic locations.
The model for such a cloud computing environment leads to
high cost when there is an increase in number of nodes. It is
also difficult to apply the load balancing strategy when the
environment is very large and complex. So cloud partitioning is
chosen. Cloud partitioning divides the public cloud into several
cloud partitions by random selection of nodes. When the
environment is very large and complex, these divisions
simplify load balancing. The Load Balancing model includes
Main Controller and Balancers, performs the load balancing
solution. The Main Controller selects the best cloud partition
and Balancers choose right load balancing strategy to distribute
the jobs to cloud partition.
The load balancing model for public cloud using cloud
partitioning concept uses a switch mechanism to choose
different strategies during different situations. The idle status
uses an improved Round Robin algorithm and normal status
uses a game theory based load balancing strategy. When the
cloud partition state is overloaded, a random node is selected as
best node to perform the load balancing, cloud partitioning is
repeated. It reduces the working efficiency and expected
response time of the system. To overcome this issue we present
an approach to develop a novel load balancing strategy for
overloaded cloud partition.
3.1 Load Balancing Strategy for Overloaded
Partition
It is evident that the working efficiency of cloud computing
environment decreases when the cloud partition status is
overloaded. So a novel load balancing model is proposed to
avoid this problem by incorporating two queues. Figure 3.1
depicts the design of the cloud architecture for this approach.
According to this design each node maintains two queues,
Priority queue and Non priority queue. Priority queue is used
when the cloud partition status is idle or normal and Non
priority queue is used when partition status is overloaded.
When user submit job request job allocation is performed
either by load balancing approach or by scheduling algorithm.
When a new job arrives, if cloud partition status is idle or
normal then load balancing approach is used to select a best
node for executing the job. The load balancing approach uses
priority queue for all arriving jobs. The jobs assigned to the
nodes have the same priority and total CPU power is shared by
all the jobs in the queue. When the cloud partition status is
overloaded then queue is splitted into two. Then the jobs in idle
and normal partition status are moved to Priority queue and the
jobs after overloaded status are moved to non-priority queue. A
separate scheduling algorithm is used for this allocation to Non
priority queue.

__________________________________________________________________________________________
ARCHITECTURE OF THE SYSTEM
Fig 3.1 System Architecture
The overloaded load balancing model is a modified approach
of existing load balancing model. Fig 3.2 shows the overall job
assignment strategy for overloaded cloud partition. In this
model when a new job arrives the main controller chooses best
cloud partition for the arriving job. The cloud partition status is
then evaluated. When the partition status is idle and normal the
jobs are assigned to the nodes according to the existing load
balancing strategy. The Idle cloud partition status uses
improved round robin algorithm and Normal partition status
uses game theory based load balancing strategy.
When an overloaded condition occurs the existing model
selects a random node as best node to perform load balancing
and cloud partitioning is repeated. It reduces the working
efficiency and expected response time of the system. So
overloaded load balancing model is used to avoid this problem
by maintaining two queues, Priority and Non priority queue.
At overloaded condition the jobs in idle and normal cloud
partition are moved to Priority queue and the jobs after
overloaded status are moved to non-priority queue. A separate
overloaded partition scheduling algorithm is used for this
allocation to non priority queue. When a priory job ends, then
one of the jobs in non priority queue moves to priority based on
arrival time and processing power required. The modified job
assignment strategy for overloaded cloud partition is shown
below.
At overloaded condition node provides X% of its CPU power
to priority queue and 1-X% to Non priority queue. For example
if X=75% and if there are three jobs in priority queue then 75%
of CPU power is shared by all three jobs in priority queue.
Initially X=100% so there is no Non priority queue. When an
overloaded situation occurs, queue is splitted into two. The
node gradually increases CPU power in non priority queue and
decreases CPU power in priority queue i.e. node provides a
threshold of 75% CPU power to Priority queue and 25% to
jobs in Non priority queue. This threshold value is calculated
using overloaded partition scheduling algorithm. The main
benefit of the overloaded load balancing model lies in response
time and fault tolerance. It further improves the efficiency by
considering all the jobs at overloaded status.
Fig 3.2 Job assignment strategy for overloaded status
3.2 Scheduling Algorithm
Priority Queue
Non Priority Queue
Node
Load
Balancing
Scheduling
Algorithm
Job Allocation User
Submit Job
Request
End
Job arrive at Cloud
Partition Balancer
Assign jobs to nodes
according to strategy
Each node maintains two
queues
Job arrive at Main
Controller
Choose Cloud Partition
Cloud
Partition
State
Start
Jobs already in queue
moves to priority queue
Jobs after overloaded
moves to non priority
queue
Apply Scheduling algorithm
for allocation
Overloaded
Idle or normal

__________________________________________________________________________________________
Scheduling algorithm is used for the allocation of jobs to non
priority queue when overloaded situation occurs. When cloud
partition status become overloaded, each node maintains two
queues Priority queue and Non priority queue. The priority
queue contains jobs when the cloud partition status is idle and
normal. Here the total CPU power is shared by all jobs and
equal priority is set for all jobs. When overloaded occurs queue
is splitted into two. The jobs already in the queue moves to
priority queue and jobs after overloaded condition moves to
non priority queue. Scheduling algorithm is used for this
allocation of jobs to non priority queue. In the scheduling
algorithm the jobs are splitted according to a threshold. The
threshold value is set as 75% and 25% for Priority queue and
Non priority queue when cloud partition status is overloaded.
At the initial stage 100% CPU power is shared by all the jobs
in priority queue. When overloaded occurs 95% CPU power is
given to jobs in priority queue and 5% CPU power is given to
jobs in non priority queue. As jobs in non priority queue
increases the CPU power given to them increases up to
threshold value of 25% and CPU power given to jobs in
priority queue decreases up to threshold value of 75%. The
threshold value is calculated using Job Splitting Index.
The Job splitting index is calculated for all the nodes in each
partition. Then calculate the average Job splitting index from
the node Job splitting index value.
When overload occurs the threshold value is set as 75% and
25% for jobs in priority and non priority queue respectively.
Allocation of job is done by selecting the best partition with
minimum average job splitting index value. For each partition
Pi select a partition if avg(Pi) < min. The selected partition is
then assigned to partition controller. At partition controller the
job is assigned to the node with minimum avg value.
Overloaded Partition Scheduling Algorithm
For each node i
For each partition Pi
Calculate Job splitting index of each node
Calculate the average Job splitting index for each partition Pi
1. Allocation of job at Main controller
Main controller chooses the best partition with min avg
Job splitting index value.
Min=α
For each partition Pi
If avg(Pi) < Min
Selected= Pi
End if
End for
2. Allocation of job at Partition controller
On arrival of job the partition controller allocate job to
the node with min Job splitting index value
4. PERFORMANCE ANALYSIS
The resulting load balancing model has been implemented and
a graph has been plotted. The graph 4.1 shows the comparative
performance of the response time of existing and overloaded
load balancing model, with the Y axis showing the effect of
improved response time on increased number of jobs in X axis.
This graph demonstrates that overloaded load balancing model
performs well as number of jobs increases. As the number of
jobs increases the proposed overloaded load balancing model
provides minimum response time.
100
90
80
70
60 Existing
50 Model
40 Proposed
30 system
20
10
0
100
12
5 150
17
5 200225
25
0 30050 75
Fig 4.1 Number of Jobs against Response time
The graph 4.2 shows the comparative performance of the fault
tolerance of existing and overloaded load balancing model,

__________________________________________________________________________________________
with the Y axis showing the effect of improved fault tolerance
on increased number of jobs in X axis. The overloaded load
balancing model provides better fault toleranace when the
number of jobs increases.
30
25
20
15 Existing
10
Model
Proposed
5
Model
0
75 100
12
5 150175
20
0 225250
30
050
Fig 4.2 Number of Jobs against Fault tolerance
From the graph plotted it is proved that our overloaded load
balancing model minimizes the load balancing in cloud
environment and there by increases overall performance of the
cloud system. The proposed model is fault tolerant when
number of jobs increases
5. CONCLUSIONS
This is a modified approach of load balancing model aimed at
the public cloud which has numerous nodes with distributed
computing resources in many different geographic locations
with the main aim of load balancing of nodes. The main benefit
of this approach lies in the development of a load balancing
strategy for overloaded cloud partition. When overloaded
condition occurs the jobs in idle and normal partition status are
moved to non-priority queue. An overloaded partition
scheduling algorithm is used for this allocation to Non priority
queue. When a priory job ends, then one of the jobs in Non
priority queue moves to priority based on arrival time and
processing power required.
ACKNOWLEDGEMENTS
I would like to express my gratitude to all those who gave me
help to complete this project. A special thanks to my guide
Prof. P Mohamed Shameem, H.O.D., CSE, TKM Institute of
Technology. I am also thankful to staffs of the institution for
guiding and providing me superior computing facilities. Last
but not least I would like to thank almighty for making this
project a reality.
REFERENCES
[1] N. G. Shivaratri, P. Krueger, and M. Singhal, “Load
distributing for locally distributed systems”,
Computer,vol. 25, no. 12, 1992.
[2] B.P Rima, E.Choi, and I.Lumb, “A Taxonomy and
Survey of Cloud Computing Systems”, Proceedings of
5th
IEEE International Joint Conference on INC,IMS
and IDC,Seoul,Korea, 2009.
[3] B P Gaochao, “Load balancing model based on cloud
partitioning for the public cloud”, IEEE transactions on
cloud computing, 2013.
[4] M. Randles, D. Lamb, and A. Taleb-Bendiab, ” A
comparative study into distributed load balancing
algorithms for cloud computing”, IEEE 24th
International Conference on Advanced Information
Networking and Applications, Perth, Australia, 2010
[5] K. Nishant, P. Sharma, V. Krishna, C. Gupta, K. P.
Singh, N. Nitin, and R. Rastogi, “ Load balancing of
nodes in cloud using ant colony optimization” 14th
International Conference on Computer Modelling and
Simulation (UKSim), Cambridge shire, United
Kingdom, 2012
[6] Nidhi Jain Kansal1, Inderveer Chana, “Cloud Load
Balancing Techniques: A Step Towards Green
Computing”, IJCSI International Journal of Computer
Science, 2012.
[7] Shantanu Dutt, “New Faster Kernighan-Lin-Type
Graph Partitioning Algorithms”, IEEE ,1993
[8] Tarun Kumar Ghosh, Rajmohan Goswami, “Load
Balanced Static Grid Scheduling Using Max-Min
Heuristic” , 2nd IEEE International Conference on
Parallel, Distributed and Grid Computing, 2012.
[9] Zehua Zhang, Xuejie Zhang, “A Load balancing
mechanism based on Ant colony and complex network
theory in Open Cloud Computing Federation”, 2nd
International Conference on Industrial Mechanism and
Automation,2010.
[10] Z. Chaczko, V. Mahadevan, S. Aslanzadeh, and C.
Mcdermid, “Availability and load balancing in cloud
computing”, International Conference on Computer
and Software Modeling, Singapore, 2011.
[11] Daniel Grosu, Anthony T.Chronopoulos,” A game
theoretic model and algorithm for load balancing in
distributed systems”, 16th
International Parallel and
Distributed Processing Symposium, 2002.
[12] Gowtham Gajala, “Cloud Computing: A State of Art of
the Cloud”, International Journal of Computer Trends
and Technology, 2013.
[13] Syed Tauhid Zuhori,Tamana Shamrin,Runia
Tanbin,Firoz Mahmud, “ An Efficient Load Balancing
approach in cloud environment by using Round Robin
algorithm” , International Journal of Artificial
Intelligence and Mechatronics, 2013.
[14] Rashmi K.S, Suma.V, Vaidehi.M, “Enhanced load
balancing approach to avoid deadlocks in cloud” ,

__________________________________________________________________________________________
International Journal of Computer Application on
Advanced Computing and Communication
Technologies for HPC Applications, June 2012.
[15] Tejinder Sharma, Vijay Kumar Banga, “Efficient and
Enhanced Algorithm in Cloud Computing”,
International Journal of Soft Computing and
Engineering, March 2013.
[16] Marios D. Dikaiakos and George Pallis, Dimitrios
atsaros, Pankaj Mehra, Athena Vakali, “Cloud
Computing: Distributed Internet Computing for IT and
Scientifc Research”, IEEE 2009.
[17] Yi Lu, Qiaomin Xie, Gabriel Kliot, Alan Geller, James
R. Larus, Albert Greenberg, “Join-Idle-Queue: A Novel
Load Balancing Algorithm for Dynamically Scalable
Web Services”.
[18] Rade Stanojevi´c, Robert Shorten, “Load balancing vs.
distributed rate limiting: a unifying framework for
cloud control”.
[19] Hao Liu, Shijun Liu, Xiangxu Meng, Chengwei Yang,
Yong Zhang, “LBVS: A Load Balancing Strategy for
Virtual Storage”, International Conference on Service
Sciences, 2010.
[20] Che-Lun Hung, Hsiao-hsi Wang and Yu-Chen Hu,
“Efficient Load Balancing Algorithm for Cloud
Computing Network”.
[21] Yi Zhao, Wenlong Huang, “Adaptive Distributed Load
Balancing Algorithm based on Live Migration of
Virtual Machines in Cloud”, Fifth International Joint
Conference on INC, IMS and IDC, 2009.
[22] Vlad Nae, Radu Prodan, Thomas Fahringer, “Cost-
Efficient Hosting and Load Balancing of Massively
Multiplayer Online Games”, IEEE 2010.
[23] Aameek Singh, Madhukar Korupolu, Dushmanta
Mohapatra, “Server-Storage Virtualization: Integration
and Load Balancing in Data Centers”.
[24] Shu-Ching Wang, Kuo-Qin, Wen-Pin Liao and Shun
Sheng Wang, “Towards a Load Balancing in a Three-
level Cloud Computing Network”, IEEE, 2010.
[25] Amandeep Kaur Sidhu, Supriya Kinger, “Analysis of
Load Balancing Techniques in Cloud Computing”,
International Journal of Computers & Technology,
April 2013.
[26] Jeffrey M. Galloway, Karl L. Smith, Susan S. Vrbsky,
“Power Aware Load Balancing for Cloud Computing”,
Proceedings of the World Congress on Engineering and
Computer Science, 2011.
[27] Yang Xu, Lei Wu, Liying Guo, Zheng Chen, “An
Intelligent Load BalancingAlgorithm towards Efficient
Cloud Computing”, AI for Data Center Management
and Cloud Computing: Papers from the AAAI
Workshop, 2011.
[28] Ratan Mishra and Anant Jaiswal, “Ant colony
Optimization: A Solution of Load balancing in Cloud”,
International Journal of Web & Semantic Technology
(IJWesT), April 2012.
[29] H K Sawant,Sachin Shelke, “A Non cooperative
approach for non cooperative load balancing in
distributed systems”, Journal of Information,
knowledge and Research in Information Technology.
[30] Md.Firoj Ali, Rafiqul Zaman Khan, “The Study on
Load Balancing Strategies in Distributed Computing
System”, International Journal of Computer Science
and Engineering Survey, April 2012.

A novel load balancing model for overloaded cloud

More Related Content

What's hot (18)

Viewers also liked (20)

Similar to A novel load balancing model for overloaded cloud (20)

More from eSAT Publishing House (20)

Recently uploaded (20)

A novel load balancing model for overloaded cloud