A-Matchmaking-Strategy-Of-Mixed-Resource-On-Cloud-Computing-Environment.pdf

INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 4, ISSUE 12, DECEMBER 2015

ISSN 2277-8616

A Matchmaking Strategy Of Mixed Resource On
Cloud Computing Environment
Wisam Elshareef, Hesham A. Ali, Amira Y. Haikal
Abstract: Today, cloud computing has become a key technology for online allotment of computing resources and online storage of user data in a lower
cost, where computing resources are available all the time, over the Internet with pay per use concept. Recently, there is a growing need for resource
management strategies in a cloud computing environment that encompass both end-users satisfaction and a high job submission throughput with
appropriate scheduling. One of the major and essential issues in resource management is related to allocate incoming tasks to suitable virtual machine
(matchmaking). The main objective of this paper is to propose a matchmaking strategy between the incoming requests and various resources in the
cloud environment to satisfy the requirements of users and to load balance the workload on resources. Load Balancing is an important aspect of
resource management in a cloud computing environment. So, this paper proposes a dynamic weight active monitor (DWAM) load balance algorithm,
which allocates on the fly the incoming requests to the all available virtual machines in an efficient manner, in order to achieve better performance
parameters such as response time, processing time and resource utilization. The feasibility of the proposed algorithm is analyzed using Cloudsim
simulator, which proves the superiority of the proposed DWAM algorithm over its counterparts in literature. Simulation results demonstrate that proposed
algorithm dramatically improves response time, data processing time and more utilized of resource compared Active monitor and VM-assign algorithms.
Index Terms: Cloud Computing; Resource management; Matchmaking; Load balance
————————————————————

1 INTRODUCTION
Cloud Computing is an emerging trend in IT environment.
Cloud computing, as figure 1 depicts, is a style of
computing that involves on-demand access to a shared
pool of computing resources such as (network, servers,
storage, applications and services), delivering hosted
services over the Internet and storing data online, there are
so many complex calculations and concepts implemented
to achieve better and better use of resources and
performance.

Fig 1: cloud computing [1]
Cloud computing allows each user to use the software and
computing services on demand at any time, in any place
and anywhere through the Internet. Cloud computing mainly
deals with computing, software, data access and storage
services that may not require knowledge of the end-user’s
geographical location and system configuration, which is to
provide services [2]. Clouds exhibit varying demands,
system sizes, supply patterns and resources (hardware,
software, and network); users have heterogeneous,
dynamic, and Quality of Service (QoS) requirements; and
applications have varying performance, workload, and
dynamic application scaling requirements [3]. The main
objective of cloud computing is to provide easy, scalable
way to computing resources and IT services. Fast
development of cloud computing appears through many
organizations such as GoGrid, Google, Rack space,
Microsoft, Amazon EC2 cloud computing and Apple to
provide cloud services to various consumers. The cloud
system dynamically allocates computing resources for the

customer/ user in response to customers’ resource
reservation requests and in accordance with customers’
QoS requirements [4]. The characteristics of cloud
computing are multi-tenancy, rented services delivery
model, on-demand usage, external data storage,
transparent, rapid elasticity, a broad network access,
resource pooling and measured service[5, 6]. An important
part of the cloud is the resource management. The
resource management strategy in cloud should effectively
utilize the pool of resource and achieve a high system
performance. Resource management can be achieved
through some sort of load balancing among the
participating nodes. On that point are some metrics that will
serve to measure the efficiency of each load balancing
techniques (LBT). LBT in a cloud environment; consider
various parameters [7, 8] such as response time,
throughput, scalability, reliability, QoS, resource utilization
and fault tolerance. Resource management affects three
basic criteria for system evaluation, They are the
performance, functionality and cost. Inefficient management
of resources has a direct negative effect on performance
and cost. You can also indirectly affect system functionality
[9]. Matchmaking and scheduling are important issues
performed by resource managers in the cloud [10].
Resource allocation in a cloud environment involves two
phases, matchmaking is the first phase and scheduling is
the second phase. Matchmaking is defined as, the method
of allocating jobs associated with user requests to
resources designated from the obtainable resource
pool..Load balance means distribute load of multiple
resources to achieve maximum throughput, minimize the
response time and to avoid the overloading at a certain
node. Both matchmaking and scheduling need to satisfy
users’ QoS requirements defined in a service level
agreement (SLA).

1.1 Motivation
Cloud computing is a computing paradigm that can provide
on demand, dynamic and scalable virtual resources through
the Internet service to users. These resources are
considered the backbone of the cloud. Recently, there are
several techniques for the management of these resources,
including matchmaking which means map incoming request
216

IJSTR©2015
www.ijstr.org

INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 4, ISSUE 12, DECEMBER 2015

to different resource, and thus exposed to the problems of
satisfying the requirements of users, better utilization of
different types of resources and dividing the load equally to
maximize the throughput and minimize response time. Load
balancing is required to distribute the dynamic local
workload evenly across all the nodes. It helps to achieve
user satisfaction and maximize the resource utilization ratio
by ensuring an efficient and equitably for all computing
resources allocation.. The important issue here is how to
achieve load balance on VM and maintain QoS so that
satisfying the requirements of users. Although literature is
plentiful with huge number of researches that provide
numerous load balancing strategies, there is still a crucial
need for an efficient load balancing strategy to satisfy user
requirements and simultaneously maximizing resource
utilization. In this paper, we will propose a new technique
for
matchmaking
between
arrival
request
and
heterogeneous resources in cloud environment. The
innovation of the proposed technique is to distribute load
between the heterogeneous resources, and on the other
hand to satisfy the requirements of end-users. Without load
balancing, users could experience delays, timeouts and
possible long system responses, and our goal is to improve
performance metrics such as response time, processing
time, resource utilization and avoid overload. The rest of
this paper is arranged as follows. Section 2 explores cloud
computing challenges and emphasizes on resource
management challenges. Section 3 related work is
discussed. Section 4 the proposed technique. Section 5
experimental result. Finally, we present some concluding
remarks.

2. CLOUD COMPUTING CHALLENGES
This section lists the key research issues and articulate
future research directions that have not boon fully
addressed related to cloud computing arena. Some of the
research challenges in cloud computing can be categorized
as shown in Figure 2.

ISSN 2277-8616

TABLE 1
CLOUD COMPUTING RESOURCE MANAGEMENT PREVIOUS
Resource
management
strategy

Pros

Cons

Linear
1 Scheduling
Strategy [11]

Improved throughput
Response time
Improved
resource
utilization

Not
suitable
interactive real
application

PRE-Copy
Approach for
2 Scheduling
[12]

Page level
hardware

Long forwarding chains
Delayed
user
experiences

3

Matchmaking
and
Scheduling
[10]

Just-In-Time
4 Resource
allocation [13]
5

MiyakoDori
[14]

A Two Tired
On- Demand
Resource
Allocation
6
Mechanism for
VMBased
Data
Centers[15]

protection

Less the delay
Economic

Cost effective
Memory reuse
Shorter migration time
It
addresses
the
problems of availability
and scalability.
If a failure of overall
resource
allocation
occurs, then the local
resource
assignment
will work conversely.So
no failure of resource
allocation is occurred

for
time

The
uncertainties
associated with such
type of ―matchmaking‖.
Unequal distribution of
the load on the various
resources.
Lack
of
knowledge
regarding local resource
management policies.
Prediction error
Use of recursive data
structures
Efficient only in cases
where migration back to
the same system

Application
workload
scheduling
is
not
considered.
Mismatch between the
on demand resource
and workload dispatch.

Given the table 1 there some problem in matchmaking that
is unequal distribution of the load on the various resources,
strive to overcome them to submit our proposal. In
literature, there are many existing load balancing
techniques that mainly focus on reducing associated
overhead, service response time and improving
performance. Table 2 lists some of these techniques in
conjunction with the advantages and disadvantages of each
technique.

Fig 2., Taxonomy of challenges in cloud computing
Resource management is considered one of the Vertis
research area in the cloud environment. The previous
efforts in cloud computing resource management can be
summarized as presented in table 1 with the advantages
and disadvantages of each approach.
217
IJSTR©2015
www.ijstr.org

INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 4, ISSUE 12, DECEMBER 2015



TABLE 2:
LOAD BALANCING TECHNIQUE OF CLOUD COMPUTING
Load
Balancing
Technique

Pros

Cons

Equality
distribution
workloads for all
the nodes

Job
processing
time
is
not
considered.
Decrease
Resource
Utilization
Does not save the
state
of
the
previous allocation
of a VM to a
request

Round Robin[16] [17]

1

2

3

4

5

6

The first request is
allocated
to
a
randomly picked VM.
Subsequent requests
They are assigned
Circular order.
Round Robin with
Server Affinity: A VM
Load Balancing [18]
save the state of the
previous allocation of
a VM to a request and
VM
state
(available/busy)
Active
VM
Load
Balancer [20]
Maintained number of
requests
currently
allocated to each VM.
Request is allocated
to the least loaded
VM.
Weighted
Active
Monitoring LB
Algorithm [16]
An assigned weight to
VMs.
Task
is
assigned to the least
loaded
and
the
highest weight VM.
ESCE Algorithm [19]
If
there
is
an
overloaded VM then
distributes some of
the tasks to some idle
VM
Throttled
load
balancer [18]
Maintained the state
of each VM.
Request is accepted if
found in the table
otherwise the request
is queued

Improved
Response time
Improved
Data
center processing
time
Compared with
Round robin

It is not clear for
the best use of the
resources
and
utilize them.

Request
is
allocated to the
least loaded VM.

Processing power
and
capacity
hardware of VM is
not considered.

Considers
both
the load Weight of
available VMs.
Increase
response
time
and
processing
time

It did not take into
account resource
utilization.

Improver
response
time
and
data
processes time

Not fault tolerant
because of single
point of failure.

tries to distribute
the load evenly
among the VMs.
Response
time
improved

other parameters
are not taken into
account such as:
weight of VM,
processing time,
etc

Given some of the problems of previous strategies as in the
table 2, we strive to overcome some of the problems by
submitting a proposal, which improves the performance
metric. Problems such as processing power of VM and
resource utilization, that means optimal use of resources.
Based on the current status of the system, the load
balancing algorithms can be divided into two categories as
static and dynamic load balancing algorithms:
 Static load balance algorithms: In this algorithm
distributed the traffic equally between servers. By this
approach, decide how to distribute the workload
accordance with prior knowledge of the problem and
characteristics of the system. These algorithms such as
Round Robin, Ant Colony optimize, Threshold and
Randomized algorithms.

ISSN 2277-8616

Dynamic load balance algorithms: In this algorithm
does not depend prior knowledge of the problem and
characteristics of the system. Dynamic algorithms use
status information to make decisions during the
execution of the program. An important advantage of
this approach is that its decision for balancing load
based on the current status of the system. These
algorithms such as Min-Min, Central Queue, Active
Monitor load balancer and Throttled load balancer
algorithms.

The proposed algorithm based on the second type of load
balance algorithms that is a dynamic load balance algorithm,
specifically conducting improvement on Active Monitor load
balancer algorithm.

3. RELATED WORK
Matchmaking has received considerable attention from
researches mostly in the context of on-demand requests.
Many studies and analysis have been performed on
matchmaking and load balancing for the cloud environment.
The emerging needs and loads in cloud environment had
driven research community to developed various load
balancing strategies. Shikharesh Majumdar et al [30]
proposed Any-Schedulability Criterion to deal with problem
of matchmaking in an environment that comprise opaque
resources that its local schedule policies are not know. In
this environment, deal with advance reservation requests,
each request has an earliest start time and deadline.
However, the proposed criterion did not take into account the
priority utilization of resources. Jasmin James et al [16]
proposed weighted active monitoring load balancing
algorithm as an improvement over the Active VM Load
Balancer [20] via assigning a weight for each VM.
Experimental result showed that their proposed algorithm
achieved better processing time and response time;
however, the proposed algorithm didn't consider process
duration for each individual request. Komal Mahajan et al
[18] deployed Round Robin (RR) approach [25] for VM Load
Balancing. Their proposed algorithm introduced an
improvement over the RR algorithm, as it included the state
of previous allocation of VM to a request, as a result of
experiences that gives better results than the RR algorithm.
Tejinder Sharma et al. [21] proposed an enhanced load
balance algorithm which lives migration of load is done in a
virtual machine to avoid the under utilization and hence
improving data transfer cost, this algorithm adopted on round
robin and throttled algorithms to improve performance metric
such as data processing time and response time, but some
variables are not taken into account such as weight of VM.
Zaouch et.al [22] presented a study about load balancing
techniques in the cloud computing and now these affects
Qos. Domanal et.al [23] proposed VM-assign Load Balance
algorithm. This algorithm specifies the incoming requests to
the available resources. Their proposed algorithm is a
modified version of Active Monitoring Load Balance
algorithm (AMLB)[20] which maintains information about all
VMs and number of current requests allocated to VM. When
a new request arrives, load balancer identified least load VM
by id. Load balancer return VM id to data center controller.
Data center controller sends the request to VM identified by
id. Data center notify load balancer of the new allocation. In
VM-assign algorithm the first allocation of VM is similar to
218

IJSTR©2015
www.ijstr.org

INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 4, ISSUE 12, DECEMBER 2015

AMLB but next allocation put the condition that is not used in
previous assignments, however, the least loaded VM which
will not chosen in the next iteration may have good
processing speed[17]. VM-assign algorithm chooses the VM
without the knowledge of his processes power that can do
the job or not. Shahapure et.al [24] proposed time sliced and
priority load balancing algorithm, it's based on the principle
of time scheduling and priority of requests and it considered
an enhancement round robin algorithm. Experimental results
showed that the algorithm reduced the waiting time and
turnaround time. Rashmi et.al [26] proposed Shortest Job
First Scheduling with threshold (SJFST) algorithm, using
various threshold values, which is considered as a Timer
determines the period for the execution of the job.
Experimental results showed that the algorithm reduced job
rejection rate compared Shortest Job First Scheduling
(SJFS) Pan, J.-S. et al [27] proposed Interaction Artificial
Bee Colony (IABC) load balance algorithm, it's based on
principle of task scheduling to VM.Proposed algorithm
makes all task scheduling by using parameter value
calculated according to gravity formulation compared to
Artificial Bee Colony (ABC) [28], which used this parameter
random number during 0 to 1. Experimental results prove
that IABC is more efficient compared to ABC. Zhan, Z.-H. et
al [29] Proposed a load balance aware genetic algorithm
(LAGA) with Min-min and Max-min to solve task scheduling
problems, so that used Time Load Balance (TLB) strategy to
help establish the fitness function with makespan.
Experimental results prove the LAGA algorithm improved
several task-scheduling problems compared with another
algorithm that not used TLB.

3.1 problem formulation
With the ever increasing number of incoming requests of
users, there is an urgent need to matchmaking requests for
resources, in order to matchmaking required distributed the
workload evenly between the different resources, which
causes some of resources not use the more than the others,
which means increasing the utilization of resource. That
necessitates achieve load balancing between those
resources. However, there are only a limited number of VM
load balancing having improved the performance, such as
response time and resource utilization. Accordingly, this
paper proposes new strategy to achieve load balance
between
heterogeneous
resources
and
improve
performance compared to previous strategies in this area.

4. PROPOSED MATCHMAKING FRAMEWORK
Figure 3 depicts the proposed layered model for cloud
computing matchmaking. The proposed framework consists
of three layers namely, user layer, core middleware layer and
system level layer. Each layer performs a specific function
and consists of different module. The innovation of the
proposed framework is using load balancing algorithm to
achieve user’s Qos requirements and better resource
utilization.

ISSN 2277-8616

Fig3.Proposed Framework of Cloud Computing
The components of the framework are as follows:
1. The user layer: this layer consists of two sub layers as
follows:
 User level: is used by the user to deal with the
services provided by cloud.
 User level middleware: This component provides
environments
and
tools
simplifying
the
development and the deployment of applications in
the Cloud and is constitutes the access point of
applications to the Cloud.
 Core middleware layer: This component is
responsible for providing a suitable run-time
environment for applications and to exploit the
physical resources. This layer consists of:
 Broker: This component is responsible for
interaction with clients and understanding their
application needs. It performs discovery and
classification of appropriate services using other
components such as the Ranking systems. The
cloud broker contains:
 SLA Management:
Is the component that keeps track of customers’
SLAs with cloud providers and their fulfillment
history.
 SMI Calculator:
Is the component that calculates the various Key
Performance Indicators (KPI’s) which they are
used by the classification system to prioritize cloud
services.
 Monitoring: this component first discovers Cloud
services that can meet users essential QoS
requirements. Then, it monitors the performance of
the Cloud services. It also keeps track of the formal
requirements of SLA previous customers are being
satisfied by the Cloud provider.
 Service Catalogue: Catalog: stores the services
and their characteristics advertised by various
Cloud providers.
 System level layer: This component is
characterized by the physical resources such as
clusters, datacenters, and spare desktop
machines.
219

IJSTR©2015
www.ijstr.org

INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 4, ISSUE 12, DECEMBER 2015

4.1 The proposed DWAM Algorithm
Some load balancing technique distributes the load among
all nodes without node configuration. The Proposed
framework will distribute the load with node configuration like
based on weight allocated to the server node. Form 4 it
explains the mechanism of load balance algorithm's action in
cloud computing, each load balance algorithm has load
balancer that identifies VM and send VM IDs to the data
center controller, which is responsible for management
allocate request to VM.
Cloudlet1
11

Cloudlet2
2

Cloudlet
m

Data Center Controller
Load balancer

Matchmaking
Techniques/ Algorithm
used

ISSN 2277-8616

F
= Finish time of user request
A
= Arrival time of user request
Ds
= data size of single request
Bw
= Band width
Nlat = Network latency
WCnet = Weight constant of network
Here the weight constant assumption parameter into scale
between 0 and 1; divide to CPU, Memory and Network, so
that the total is currently just 1. The CPU has greater impact
on the execution of a VM comparing in memory or net. So, it
has a maximum weight constant. The use of a constant
value in the equation, because there are different units of
measurement for each of memory, CPU and network.
Proposed DWAM algorithm VM load balancer and the
related flowchart are given in Figures 5 and 6 respectively.
Main ideas, that upon the proposed DWAM algorithm are
assigned weight in a dynamical way by equation for each
VM, depending on process completion time (process
duration), a VM with least load and observance if VM with
least load used in the last iteration.

Virtual Machine Manager

Vm -n

Vm -1

Fig 4: Load Balance Algorithms Execution
Dynamic Weight Active Monitor (DWAM) Load Balance
algorithm is a modification for the VM-assign Load Balance
[23] and Active Monitoring Load Balancer [20] by assigning a
dynamic weight to each VM. Unlike previous algorithms
which calculated weight in a static way. The proposed
DWAM algorithm introduces the concept of dynamic weights
with active monitoring. Each VM is assigned a dynamic
weight and according to the highest weight, they receive
more connections. In a situation, when all the weights
become equal, VM will receive balanced traffic. The VM is
assigned a varying amount of the available processing
power of VMs. To these VMs of different processing powers;
the requests are allocated to the most powerful VM and then
to the lower and so according to their weight and its
availability. Hence, optimizing the given performance
parameters. After selecting the VM with the least load and
with least process duration is identified, so that if VM was
used the last time we choose another, it allocates requests
to the most powerful VM according to the weight assigned.
The main objective of the proposed algorithm is to achieve
better response time, processing time and resource
utilization. The weight of VM is calculated as shown in
equation 1:
VMw = Cs ∗ (1 − CPUut ) ∗ WCcpu + (1 − (
(F − A +

Ds
BW

+ Nlat ) ∗ WCnet

Tm U m
)
Tm

∗ WCm ) −

(1)

Where
Cs
= Clock speed
CPUuti = CPU Utilization
WCcpu = Weight constant of CPU
Tm
= Total Memory
Um
= Used Memory
WCm = Weight constant of memory
220
IJSTR©2015
www.ijstr.org

INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 4, ISSUE 12, DECEMBER 2015

DWAM

ISSN 2277-8616

algorithm

Input: No of incoming request r1, r2,………rn.
Available VM vm1, vm2,…… vmn.

Output: Mapping request of VM’s such that Qos
parameters are fulfill and achieve

load balance.

Steps:
Create VM’s on data center with appropriate
memory, storage, bandwidth, ect.
2. Calculate weight factor (Wi) for all VM according to
equation (1).
3. DynamicWeightActiveVMLoadBalancer maintains
an index table of VMs, associated weighted count
and the number of requests currently allocated to the
VM. At start all VM's have 0 allocations.
4. When requests arrive at the data center it passes to
the load balancer, it parses the table; least loaded
VM and with least process duration is selected for
execution.
Case I: if found
Check whether the chosen least
loaded VM is used immediately in the
last iteration
If YES
goto step 4 to find next least VM
If NO
Least loaded VM is chosen
5. Identifying the least loaded VM with least process
duration, then load balancer return VM id to data
center.
6. Data center send request to VM (identified by id)
and it notify the load balancer about new allocation.
7. Load balancer updates the table increasing the
allocate count for that VM.
8. When VM finishing processing the request, data
center notify the load balancer of VM de-allocate.
9. Load balancer updates the table by decreasing the
1.

allocation count for the VM by one.
10. Continue from step 4 for the next request.

Fig 5. DWAM proposed algorithm

Fig 6. Flowchart for proposed algorithm

5. EXPERIMENTAL
The proposed algorithm is implemented in cloudsim
simulation toolkit. Java language is used for implement new
DWAM load balancing algorithm. To analyze result and
compare them with existing algorithm, we used CloudAnalyst
tool for simulating proposed algorithm.

221
IJSTR©2015
www.ijstr.org

INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 4, ISSUE 12, DECEMBER 2015

5.1 Experimental setup
Cloud Analyst tool gives the real time scenario with six
geographical locations. The reservation data centers at the
same time. The users of each continent the assumed that a
small percentage of the total Internet users is during peak
hours and off-Peak hours, users are ten peak hours. For
experimentation Internet users in six different continents
considered six user bases and peak hours and off- peak
hours are given a in Table 3. We have studied Internet users
in different continents from month of May 2013.

The cloud environment set up generated was having
following configuration; data centers configuration as
presented in table 5 .Each data center has one VM. We used
five VM in the experimental and these have four-type
different configuration as in table 6.
TABLE 5:
DATA CENTERS CONFIGURATION

TABLE 3:
SIMULATION CONFIGURATION

S.No

User
Base

Region

Simultaneous
Online Users
During
Peak
Hrs

Simultaneous
Online
Users
During Off- peak
Hrs

1

UB1

0-N. America

470000

80000

2

UB2

1-S. America

600000

110000

3

UB3

2- Europe

350000

65000

4

UB4

3- Asia

800000

125000

5

UB5

4- Africa

115000

12000

6

UB6

5- Oceania

150000

30500

The number of host and data center, storage of host and VM
given in the Table4.
TABLE 4:
SIMULATION CONSIDERATION
Parameter

Value

Simulation toolkit

Cloudsim

Number of host

011

Number of Datacenter

5

Host storage

100GB

VM storage

10GB

ISSN 2277-8616

Parameter

Value

Data Center OS

Linux

Data Center Architecture

X64

TABLE 6:
VM CONFIGURATIONS
CPU
No.
processors

Of Memory

Storage
space

Bandwidth

1

1 GB

1TB

1000000 bps

2

4 GB

2 TB

1000000 bps

4

8 GB

4 TB

1000000 bps

8

16 GB

8 TB

1000000 bps

5.2 Result and analysis
Here compared the proposed DWAM load balance
algorithm with Active Monitor and VM- assign load balance
algorithms. As shown in table 7, fig 8, fig 9 the response
time is significantly better than other algorithms. From table
7 can be seen maximum response time decrease from
249.6 to 203.6, that means 18.4% improvement of
response time compared VM- assign algorithm and 44.8%
improvement compared Active Monitor algorithm. Due to
the assigned weights to each VM, which means the request
are assigned to most powerful VM and with least process
duration. In addition, improvement appears dramatically
compared Active Monitor algorithm, because it chooses VM
based only with least load.
TABLE 7:
RESPONSE TIME (MS) RESULTS

Application is deployed in five Data centers located in
different parts of the world (six different regions: R0, R1, R2,
R3, R4 and R5) as figure 7.

VM#

ACTIVE
MONITOR

VM- ASSIGN

DWAM

VM0

369.36

236.34

90.54

VM1

362.93

249.6

75.6

VM2

339.64

124.51

168.64

VM3

278.19

181.69

203.6

VM4

300.96

221.68

193.5

MAX

369.36

249.6

203.6

MIN

278.19

124.51

75.6

AVG

1651.08

1013.82

731.88

Fig 7. Data centers in different regions
222
IJSTR©2015
www.ijstr.org

INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 4, ISSUE 12, DECEMBER 2015

400
350
300
250
200
150
100
50
0
VM0

VM1

VM2

Active Monitor

VM3

450
400
350
300
250
200
150
100
50
0

VM4

VM- assign

ISSN 2277-8616

VM0

DWAM

VM1

VM2

Active Monitor

Fig 8. Response Time

VM3

VM4

VM- assign

DWAM

Fig 10. Resource utilization analysis

1800
1600
1400
1200
1000
800
600
400
200
0
DWAM

VM- assign

MAX

1600
1400
1200
1000
800
600
400

Active Monitor

MIN

200

AVG

0
DWAM

VM- assign

Active Monitor

MAX

AVG

Fig 9. Response Time results comparison

The proposed DWAM algorithm distributes the incoming
requests to all VM's. Therefore, it more utilization of the
resources compared to Active load balancer and VM-assign
as shown table 8, fig 10, fig 11. Proposed algorithm will not
allow the VM, which was allocated in its previous step, so
that take chance to other least load VM that achieved more
utilization of the resource. From table 8 we can note
proposed algorithm more utilized of resource by 59.7%
compared Active Monitor algorithm and 11% compared VMassign algorithm. This proves that the proposed contributed
to the improvement of resource utilization when it assigned
weight to each VM and VM with least process duration. In
additional, it ruled out the least load VM that used in last
iteration.
TABLE 8:
RESOURCE UTILIZATION RESULTS

MIN

Fig 11. Resource utilization results comparison

The results in table 9,fig 12, fig 13 show that proposed
algorithm reducing data center processing time compared
two algorithms. To adopt the proposed algorithm to
assigned weight for each VM play an important role in
improvement of data center processing time. Therefore, the
request allocated to highest weight of VM, which makes it
implement request in shortest possible time. As table 9 we
notice 17% improvement process time compared Active
Monitor algorithm and 11% compared VM-assign algorithm.
TABLE 9:
DATA PROCESS TIME (MS) RESULTS

VM#

ACTIVE MONITOR

VM- ASSIGN

DWAM

VM#

ACTIVE MONITOR

VM- ASSIGN

DWAM

VM0

071321
066321
03532
81325
067350

051321
044354
046378
61354
021386

VM0

0.965

0.9

0.7

VM1

0.854

0.765

0.64

VM2

0.4

0.7

0.8

VM3

0.48

0.58

0.44

VM4

242383
101364
287328
072327
085333

VM4

0.8

0.45

0.3

MAX

398.39

180.32

160.32

MAX

0.965

0.9

0.8

MIN

183.38

90.36

70.65

MIN

0.4

0.45

0.3

AVG

1344.9

772.91

675.48

AVG

3.499

3.395

2.88

VM1
VM2
VM3

223
IJSTR©2015
www.ijstr.org

INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 4, ISSUE 12, DECEMBER 2015

ISSN 2277-8616

performance can be noted when the algorithm run on more
number of data centers. We also conclude that proposed
DWAM load balance algorithm is best among other.

1.2
1

REFERENCES:
[1] Shroff, G., 2010. Enterprise cloud computing. Enterprise
Cloud Computing Technology, Architecture, Applications,
pp 241–242.

0.8
0.6

[2] Kalagiakos, P. & Karampelas, P., 2011. Cloud
Computing learning. 2011 5th International Conference
on Application of Information and Communication
Technologies (AICT), pp 7–11.

0.4
0.2
0
VM0

VM1

Active Monitor

VM2

VM3

VM- assign

VM4
DWAM

Fig 12. comparison analysis between proposed algorithm Vs other
algorithm

[4] Sidana, S., & Suri, B. 2013. Cloud Computing: A Review
International Journal of Advanced Computer Science
and Software Engineering, Vol. 3, issue 4, April

4
3.5
3
2.5
2
1.5
1
0.5
0
DWAM

VM- assign

MAX

MIN

[3] Vahora, S. & Patel, R., 2015. CloudSim-A Survey on VM
Management Techniques. IJARCCE International
Journal of Advanced Research in Computer and
Communication Engineering, Vol. 4, Issue 1, January,
pp 128–133.

[5] Chung, M. & Hermans, J. 2010. From Hype to Future.
KPMG's 2010 Cloud Computing Survey. KPMG.
[6] Peter, M. & Timothy, G. 2011. The NIST Definition of
Cloud Computing . National Institute of Standards and
Technology, Gaithersburg, MD, Sept.
[7] Manvi, S.S. & Shyam, G.K., 2014. Resource
management for Infrastructure as a Service (IaaS) in
cloud computing: A survey. Journal of Network and
Computer Applications, pp 424–440.

Active Monitor

AVG

Fig 13. Data Process Time comparison

The above shown tables and figures clearly indicates that
the parameters: response time, data processing time and
resource utilization are bit improvement in proposed DWAM
algorithm compared Active Monitor and VM-assign
algorithms.

6. CONCLUSION
In this paper proposed new strategy of load balance and
then implemented in cloud environment using CloudSim
toolkit. In proposed DWAM algorithm the requests are
located to the most power full VM, that is by assign the
weights dynamically for each VM. From the results in a
table 7, we can say Improvement largely on response time
compared other algorithms. According to the experiment,
we conclude that if select VM with least load, with least
process duration and highest weight for execution of
request then it also increase the performance of the system
also decrease the response time of the requests and data
center processing time. From table 8, we can say that the
proposed algorithm is more utilization of the resource.
When compared proposed DWAM algorithm to Active
monitor and VM-assign algorithms, we noticed response
time reduced, data processing time is also decrease and
more utilized of resource. The improvement in the

[8] Kansal, N. J. & Chana, I. 2012. Cloud Load Balancing
Techniques: A Step Towards Green Computing.
Environment International Journal of Advanced
Computer Science Issues, Vol. 9 (1).
[9] Sanchari, S. & Abhilash, K.V., 2014, A Survey on
Resource Management in Cloud Computing.
International Journal of Computer and Information
Technology, Vol. 5 (3) , 3887-3889
[10] Hoang, P. et al., 2014. Resource management
techniques for handling uncertainties in user estimated
job execution times. International Symposium on
Performance
Evaluation
of
Computer
and
Telecommunication Systems (SPECTS 2014).
[11] P, A.S., 2012. Linear Scheduling Strategy for Resource
Allocation in Cloud Environment. International Journal
on Cloud Computing: Services and Architecture
IJCCSA, Vol.2, No.1, February, pp 93–17.
[12] Soni, S.K. & Kapoor, R.K., 2013, A Survey on Pre-copy
Based live Migration of Virtual Machine in Cloud
Environment International Journal of Advanced
Computer Science and Software Engineering, Vol. 3,
issue 10, October.
224

IJSTR©2015
www.ijstr.org

INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 4, ISSUE 12, DECEMBER 2015

[13] Roy, N., Dubey, A. & Gokhale, A., 2011. Efficient
Autoscaling in the Cloud Using Predictive Models for
Workload Forecasting. 2011 IEEE 4th International
Conference on Cloud Computing, pp. 500–507.
[14] Akiyama, S. et al., MiyakoDori: A Memory Reusing
Mechanism for Dynamic VM Consolidation. 2012 IEEE
Fifth International Conference on Cloud Computing, pp.
606-613.
[15] Song, Y., Sun, Y. & Shi, W., 2013. A Two-Tiered OnDemand Resource Allocation Mechanism for VM-Based
Data Centers. IEEE Transactions on Services
Computing IEEE Trans. Serv. Comput., pp 116–129.
[16] James, J. & Verma, B., 2012. Efficient VM Load
Balancing Algorithm for a Cloud Computing
Environment. International Journal of Advanced
Computer Science & Engineering, Vol. 4, pp 16581663, ISSN: 0975-3397, September
[17] Shaw, S.B. & Singh, A.K., A survey on scheduling and
load balancing techniques in cloud computing
environment. 2014 International Conference on
Computer and Communication Technology (ICCCT), pp
87–95.
[18] Mahajan, K., Makroo, A. & Dahiya, D., 2013. Round
Robin with Server Affinity: A VM Load Balancing
Algorithm for Cloud Based Infrastructure. Journal of
Information Processing Systems, Vol.9, No.3,
September, pp 379–394.
[19] Mohapatra, S., Rekha, K.S. & Mohanty, S., 2013. A
Comparison of Four Popular Heuristics for Load
Balancing of Virtual Machines in Cloud Computing.
International Journal of Computer Applications IJCA,Vol.
68- No. 6, pp 33–38.

ISSN 2277-8616

[25] Mustafa, E. et al., 2015. Load Balancing Algorithms
Round-Robin (RR), Least-Connection and Least
Loaded Efficiency. International Journal of Computer
and Information Technology. Vol 04 – Issue 02, March.
pp 255-257.
[26] Srinivasan, R.K., Suma, V. & Nedu, V., 2013. An
Enhanced Load Balancing Technique for Efficient Load
Distribution in Cloud-Based IT Industries. Advances in
Intelligent Systems and Computing Intelligent
Informatics Springer, Heidelberg , Vol. 182, pp 479–485.
[27] Pan, J.-S. et al., Interaction Artificial Bee Colony Based
Load Balance Method in Cloud Computing.2015.
Advances in Intelligent Systems and Computing Genetic
and Evolutionary Computing.Vol. 329, pp 49–
57.Switzerland,Springer.
[28] Zhao, J., Lv, L. & Sun, H., Artificial Bee Colony Using
Opposition-Based Learning. 2015. Advances in
Intelligent Systems and Computing Genetic and
Evolutionary
Computing,
Vol.329,
pp
3–10.
Switzerland,Springer.
[29] Zhan, Z.-H. et al., Load Balance Aware Genetic
Algorithm
for
Task
Scheduling
in
Cloud
Computing.2014. Lecture Notes in Computer Science
Simulated Evolution and Learning, Vol. 8886. pp 644–
655. Switzerland,Springer.
[30] Melendez, J.O. & Majumdar, S., Matchmaking on
distributed systems with limited knowledge of resources.
2010. Ottawa ON: IEEE Performance Evaluation of
Computer and Telecommunication Systems (SPECTS),
2010 International Symposium on.

[20] Hemant, S.M. & Parag R.K. & Vinay, C. 2013. Load
Balancing On Cloud Data Centres. International Journal
of Advanced Computer Science and Software
Engineering, Vol. 3, issue 1, January.
[21] Sharma,T. & Banga, V.K. ,2013, Efficient and Enhanced
Algorithm in Cloud . International Journal of Advanced
Computer Science and Applications IJACSA, ISSN:
2231-2307, Vo. 3,pp 385-390, Issue-1, March.
[22] Zaouch, A. & Benabbou, F., 2015. Load Balancing for
Improved Quality of Service in the Cloud. International
Journal of Advanced Computer Science and
Applications IJACSA, Vol. 6,No.7,pp 184-189.
[23] Domanal, S.G. & Reddy, G.R.M., 2014. Optimal load
balancing in cloud computing by efficient utilization of
virtual machines. 2014 Sixth International Conference
on
Communication
Systems
and
Networks
(COMSNETS) IEEE. Bangalore
[24] Shahapure, N.H. & Jayarekha, P., 2015. Time sliced and
priority based load balancer. 2015 IEEE International
Advance Computing Conference (IACC).
225
IJSTR©2015
www.ijstr.org