Hybrid of Ant Colony Optimization and Gravitational Emulation Based Load Balancing Strategy in Cloud Computing

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 04 Issue: 07 | July -2017 www.irjet.net p-ISSN: 2395-0072
© 2017, IRJET | Impact Factor value: 5.181 | ISO 9001:2008 Certified Journal | Page 1236
Hybrid of Ant Colony Optimization and Gravitational Emulation Based
Load Balancing Strategy in Cloud Computing
Jyoti Yadav1, Dr. Sanjay Tyagi2
1M.Tech. Scholar, Department of Computer Science & Applications, Kurukshetra University, Haryana, India 2Assistant
Professor, Department of Computer Science & Applications, Kurukshetra University, Haryana, India
-------------------------------------------------------------------------***------------------------------------------------------------------------
Abstract- The distributed architecture of cloud
computing set up the resources distributively for delivering
the services to cloud consumers. In this paper, a novel hybrid
ACO and gravitation emulation based strategy considering
load balancing has been implemented for solving the load
balancing issue in cloud environment efficiently. Moreover,
this hybrid ACO-GELS algorithm uses the physics concept of
gravitational attraction between objects. GELS algorithm is
powerful for local search in searching space. CloudSim has
been used as a simulation tool for proposed hybrid load
balancing strategy. The proposed ACO-GELS algorithm has
been compared with the existing GA-GELS algorithm. It has
been compared on the basis of three important factors of
load balancing: resource utilization, makespan and load
balancing level.
Key Words: Ant Colony Optimization, Cloud Computing,
CloudSim, GELS, Gravitational Emulation, Load balancing.
I. INTRODUCTION
Internet technologies are growing rapidly and cloud
computing has become a hot topic to meet the user’s
needs. Cloud computing is a distributed computing
mechanism. The cloud based system provides services on
demand from anywhere in the world.
Cloud computing has a splendid future, even though there
are some important problems that need to be solved for
minimizing response time, minimizing cost, maximizing
throughput, etc.
Load Balancing is one of the main issues that need to be
considered in cloud computing as it plays a vital role in
cloud computing. As the name suggests, load balancing
means to distribute the workload evenly among the virtual
machines (VMs) [1].
For load balancing, some points need to be kept in mind:
communication between the nodes, expected load, stability
of different system, selection of nodes, arrangement of
system and nature of work to be transferred [2]. The main
objective of load balancing is to distribute the local
workload evenly to ensure that no VM is overloaded or
idle. Load balancing reduces the makespan and response
time and increases the utilization of resources. The basic
scenario of load balancing has been represented in fig-1:
Fig-1: Load Balancing Scenario
There are many static and dynamic load balancing
algorithms like FCFS, Round-Robin, ACO, GA, PSO, etc.
In this paper, for load balancing in VMs, a hybrid of
Gravitational Emulation Local Search (GELS) and ACO has
been proposed. ACO is an optimization algorithm which
uses basic foraging behavior of ants and uses the
pheromone values & GELS is powerful for local search and
weak for global search. It is based on the basic concept of
gravitational attraction in physics.
For simulation and analysis of the proposed hybrid ACO-
GELS algorithm, CloudSim simulation tool has been used.
The rest of the paper has been organized as follows:
Section II- Related Work
Section III- Load Balancing of VMs using ACO & GELS
Section IV- Proposed ACO-GELS Algorithm
Section V- Simulation Results and Analysis
Section VI- Conclusion
II. RELATED WORK
A cloud task scheduling based ACO approach was
presented by Medhat A. Tawfeek et. al. [3] for allocation of

incoming tasks to the VM’s in an efficient manner to
minimize the makespan. Therefore, the available resources
could be utilized optimally to achieve a high user
satisfaction and minimize the resource consumption. This
approach was performed using CloudSim simulation tool.
The simulation results show that ACO algorithm is better
than Round-Robin and FCFS algorithms. Also, the best
values of parameters have been experimentally obtained
for ACO algorithm.
An advanced reservation approach was developed to
provide users with the guaranteed quality of service (QoS)
[4]. Advanced reservation is a kind of process which allows
the resources to be allocated on the basis of increase in
number of accepted users request and the need of QoS in
Grid system. In this paper, GELS algorithm, a heuristic
method was used for scheduling and advanced reservation
of resources. The algorithm proposed was named as
GELSAR and compared with GA. The experiments showed
that GELSAR has better execution time in comparison with
GA which reduced to 50 percent. Also, it increased the jobs
reservation.
Hybrid algorithm of GA and GELS was proposed for load
balancing of VMs in cloud [5]. GEL is a local search
technique which uses the concept of gravitational
attraction. Therefore, it is powerful for local search and
feeble for global searches, while GA is inspired by natural
evaluation method for existence. A new set of strings was
generated in each generation. The crossover and mutation
operations were used in GA. GA is strong for global
searches. Cloud Analyst simulation tool has been used for
the implementation of this proposed algorithm. The
proposed GA-GELS tried to reduce the makespan and
improved the response time of VMs. Also, this algorithm
was compared with existing techniques like ACO, GA, FCFS
and guaranteed the QoS requirement of user’s request.
Particle Swarm Optimization (PSO) and Gravitational
Emulation based Hybrid approach has been presented by
Rakshanda et. al. [6], which was named as PSO-GEL
algorithm. The mechanism of PSO is inspired by swarm of
insects, bird flocks and fish schooling. Each particle
represents a feasible solution and have a position and a
velocity. In this paper, the results showed that PSO-GEL
was better than GA-GEL algorithm and proposed PSO-GEL
algorithm has less response time than GA-GEL.
III. LOAD BALANCING OF VMs USING ACO & GELS
3.1 Ant Colony Optimization Algorithm (ACO)
ACO is a random optimization technique which is inspired
from the food searching behavior of real ants. When ants
move from their nest to food, they deposit the chemical
substance called pheromone on their way [7]. An ant can
follow the trail of other ants by sensing the pheromone on
the ground. This pheromone concentration is used to find
the best path to the source. The more deposition of
pheromone leads to a positive feedback effect. And then
the pheromone value is updated and more ants follow that
path.
Two types of pheromones are used in ACO:
Foraging Pheromone (FP) is used for movement towards
overloaded nodes, and
Trailing Pheromone (TP) is used for tracing underloaded
nodes [8].
Disadvantages of this algorithm are overhead, stagnation
phenomenon and this algorithm converges to local optimal
solution.
Pseudocode of ACO can be represented as:
Initialize parameters;
while(termination criterion not satisfied)
Construct Solution;
Apply Local Search;
Global Pheromone Update;
Self-Adaptive Mechanism;
end
return best solution;
3.2 Gravitational Emulation Local Search
Algorithm (GELS)
GEL algorithm was given by Voundaris and Tesong in
1995, for searching in a search space. In 2004, more
powerful algorithm was proposed by Barry Webstar called
GELS (Gravitational Emulation Local Search Algorithm)
[4].
This algorithm is inspired from the basic concept of
gravitational attraction. The objects are pulled towards
each other due to gravity. Also, more closer the two
objects, stronger the gravitational force between them.
This algorithm introduced the concept of randomization
along the two primary parameters: velocity and gravity.
Newton’s formula of gravitational force is:
F= G m1.m2 (1)
R2
Here, m1 & m2 are mass of first & second object, G is
gravitational constant which has value 6.672 and R is
distance between two objects.
GELS algorithm includes a pointer that moves through the
search space and GELS allows movement by two methods.

In first method, a candidate solution is selected from the
neighborhood of the current response. In second method,
movements are allowed outside the neighborhood of
current response [5].
In GELS, formula in equation 1 is modified and the
gravitational force has been calculated as follows:
F= G (CU-CA) (2)
R2
The mass is replaced by difference of Current response
and Candidate response.
Here, CU= Current response
CA= Candidate response
R= Constant or may change on each iteration.
G= Gravitational Constant (6.672)
Makespan can be defined as the maximum completion
time, i.e.
Makespan = max(Finish_time [i,j] )
Finish_time[i,j] indicates the time at which task i ends on
VM j.
For VM, load balance can be obtained from eq 3 as
CPU_LB= makespan / avgET (3)
Here, avgET is the average execution time for all user
tasks.
So, the fitness function can be calculated as
Fit1=1/ CPU_LB (4)
IV. PROPOSED ACO-GELS ALGORITHM
Proposed ACO-GELS algorithm uses the best quality of ACO
and GELS algorithm. The basic steps of algorithm are given
below:
1) Initial population generation: GELS algorithm is used
to select the total population and high primary velocity is
taken into concern.
2) Force Calculation: Gravitational force of current
response and candidate response is calculated using
equation 2 and then it is added to the velocity of that
dimension of the candidate response.
3) Pheromone Updation: Global pheromone updation is
done using acceleration.
Acceleration[i][j] = Force[i][j] / mass[i]
4) Termination condition: Either primary velocity equals
to zero or maximum number of iteration is reached.
Pseudocode of proposed ACO-GELS algorithm:
begin
-Initialize the population of VM randomly
-Evaluate the fitness for each ant
-Assign a predefined starting soln as current soln
-Assign an initial velocity randomly in the dimension
-Calculate an initial vector velocity sum
-Best_soln = current soln
-while (velocity_sum != 0 OR i <= max_ITER)
Do
velocity_sum=0
Generate candidate response
Calculate Mass and Acceleration
Calculate the difference in GF between current
soln and candidate soln using eq 2
if ( fit(candidate soln) > fit(current soln) )
then
Best_soln = candidate soln
Update Velocity (v,force)
end
-Global pheromone update using Acceleration
-return best solution found
-end
V. SIMULATION RESULTS AND ANALYSIS
CloudSim has been used for simulation of proposed
algorithm. CloudSim is an extensible simulation
framework that allows experimentation and seamless
modeling of emerging cloud computing application
services. The main advantages of using CloudSim are:
(i) Time effectiveness
(ii) Flexibility and applicability
The proposed ACO-GELS algorithm has been simulated by
using three parameters: makespan, resource utilization
and load balancing level.
Makespan- Makespan can be defined as the time
difference between the start and finish time of tasks or it is
the completion time of tasks. Also the makespan must be
low for better performance [9].
Resource Utilization- Resource utilization means to serve
more users during specific operation time [10]. The more
the utilization of resources means more the balancing of
load i.e. it must be large.
Load Balancing Level- It includes the balancing level of
load on the VMs or how the load is distributed evenly on
the machines [11]. The load is distributed efficiently, if this
parameter is high.
The experimental results of ACO-GELS algorithm have
been compared with the existing GA-GELS algorithm [5] on
the basis of above three parameters and the results are

shown below in tabular form as well as graphically with
different number of tasks.
Here, the numbers of virtual machine are 5 for all the
experiments and numbers of tasks have been varied like
50, 100 and 200.
Table-1: Makespan Analysis (in ms)
Number of
Tasks
GA-GELS ACO-GELS
50 94.1 38.1
100 104.1 41.43
200 307.1 214.1
Table-1 shows the makespan metric values for GA-GELS
and proposed ACO-GELS algorithms with three different
scenarios. First simulation has been done with 50 tasks on
5 VMs, then with 100 and 200 tasks on 5 VM’s.
Fig-2: Performance analysis of ACO-GELS with existing GA-
GELS using five VMs and 50, 100 & 200 tasks
Fig-2 represents the simulation results of comparison of
makespan between ACO-GELS and GA-GELS in graphical
form. It is clear from the figure that proposed ACO-GELS
algorithm performs better than GA-GELS and reduces the
makespan even with the large number of tasks.
Table-2: Average Resource Utilization Analysis
Number of
Tasks
GA-GELS ACO-GELS
50 44 93.96
100 46.79 96.21
200 69.03 99.02
The average resource utilization rate for GA-GELS and
ACO-GELS has been shown in Table-2. The readings are
taken for different number of tasks i.e. 50, 100 and 200
with five VMs.
Proposed ACO-GELS algorithm has been compared with
GA-GELS based on average RU metric in fig-3. It shows that
ACO-GELS is better than GA-GELS for this factor in load
balancing of cloud system.
Fig-3: Performance analysis of ACO-GELS with existing GA-
GELS using five VMs and 50, 100 & 200 tasks
Table-3: Load Balancing Level Analysis
Number of
Tasks
GA-GELS ACO-GELS
50 33.84 95.19
100 42.08 96.83
200 74.35 99.01
Table-3 represents the load balancing level for GA-GELS
and proposed ACO-GELS for 50, 100 and 200 tasks using
five VMs.
Fig-4: Performance analysis of existing GA-GELS and
proposed ACO-GELS using five VMs for 50, 100 and 200
tasks
Fig-4 represents the comparison of ACO-GELS and GA-
GELS graphically for load balancing level (LBL) metric. The
0
100
200
300
400
50 100 200
Time(ms)
Number of Tasks
Makespan
GA-GELS
ACO-GELS
33.84 42.08
74.3595.19 96.83 99.01
0
100
200
50 100 200
Time(ms)
Number of Tasks
Load Balancing Level
GA-GELS ACO-GELS
0
200
50 100 200
Time(ms)
Number of Tasks
Average Resource
Utilization %
GA-GELS
ACO-GELS

results showed that proposed algorithm increases the LBL
as compared to the GA-GELS algorithm.
VI. CONCLUSION
This paper presents a novel load balancing algorithm
based on hybrid of ACO algorithm and GELS algorithm. The
proposed hybrid algorithm is named as ACO-GELS. ACO-
GELS algorithm provides the better result to solve cloud
computing load balancing problem to minimize the
makespan. GELS algorithm provides better result for local
searching and ACO is an optimization technique.
Experimental results conclude that ACO-GELS provides the
better results than the GA-GELS algorithm. It is also better
than FCFS, GA, SHC and ACO algorithms, since GA-GELS
results have shown it better than all these algorithms [5].
ACO-GELS reduced the makespan, increased the resource
utilization and also increased the load balancing level as
compared with GA-GELS.
Though priority of tasks and fault tolerance issues has not
been considered here, this may be considered in further
research. Researcher can include the priority issue and
fault tolerance issue for future research work.
REFERENCES
[1] N. Pasha, D. A. Agarwal and D. R. Rastogi, "Round
Robin Approach for VM Load Balancing Algorithm in
Cloud Computing Environment," International Journal
of Advanced Research in Computer Science and
Software Engineering, vol. 4, no. 5, pp. 34-39, May
2014.
[2] B. L. Webster, "Solving Combinatorial Optimization
Problems Using a New Algorithm Based on
Gravitational Attraction," Florida Institute of
Technology, Melbourne, USA, 2004.
[3] M. A. Tawfeek, A. El-Sisi, A. E. Keshk and F. A. Torkey,
"Cloud Task Scheduling Based on Ant Colony
Optimization," in 8th International Conference on
Computer Engineering & Systems (ICCES),IEEE, 2013.
[4] B. Barzegar, A. M. Rahmani, K. Z. far and A. Divsalar,
"Gravitational Emulation Local Search Algorithm for
Advanced Reservation and Scheduling in Grid
Computing Systems," in Fourth International
Conference on Computer Sciences and Convergence
Information Technology, IEEE, 2009.
[5] S. Dam, G. Mandal, K. Dasgupta and P. Dutta, "Genetic
Algorithm and Gravitational Emulation Based Hybrid
Load Balancing Strategy In Cloud Computing," in
Third International Conference on Computer,
Communication, Control and Information Technology
(C3IT), IEEE, 2015.
[6] Rakshanda, D. K. Garg and Vinod, "Particle Swarm
Optimization and Gravitational Emulation Based
Hybrid Load Balancing Strategy in Cloud Computing,"
International Journal for Scientific Research &
Development, vol. 4, no. 4, pp. 983-986, 2016.
[7] Y. Zhou and X. Huang, "Scheduling Workflow in Cloud
Computing Based on Ant Colony Optimization," in
Sixth International Conference on Business Intelligence
and Financial Engineering, 2013.
[8] S. Kumar, D. S. Rana and S. C. Dimri, "Fault Tolerance
and Load Balancing Algorithm in Cloud Computing,"
International Journal of Advanced Research in
Computer and Communication Engineering, vol. 4, no.
7, pp. 92-96, 2015.
[9] N. Sharma, S. Tyagi and S. Atri, "A Comparative
Analysis of Min-Min and Max-Min Algorithms based
on the Makespan Parameter," International Journal of
Advanced Research in Computer Science, vol. 8, no. 3,
pp. 1038-1041, 2017.
[10] A. Al-Shaikh, H. Khattab, A. Sharieh and A. Sleit,
"Resource Utilization in Cloud Computing as an
Optimization Problem," International Journal of
Advanced Computer Science and Applications (IJACSA),
vol. 7, no. 6, pp. 336-342, 2016.
[11] P. A. Pattanaik, S. Roy and P. K. Pattnaik,
"Performance Study of Some Dynamic Load Balancing
Algorithms in Cloud Computing Environment," in 2nd
International Conference on Signal Processing and
Integrated Networks (SPIN), 2015.

Hybrid of Ant Colony Optimization and Gravitational Emulation Based Load Balancing Strategy in Cloud Computing

More Related Content

What's hot (17)

Similar to Hybrid of Ant Colony Optimization and Gravitational Emulation Based Load Balancing Strategy in Cloud Computing (20)

More from IRJET Journal (20)

Recently uploaded (20)

Hybrid of Ant Colony Optimization and Gravitational Emulation Based Load Balancing Strategy in Cloud Computing