RHEA: Resource hypervisor and efficient allocator in cloud

Abstract

In this modern world, people are not ready to waste their time in waiting for long duration. That's why cloud computing is such an enor-mous number of fans that it can be rented and also pay per use. The cloud service provider is concern about the data owner's satisfaction in cloud usage. The main area they concentrate will be the security of the owner's data and the resource allocation as per the request. This paper explains how the resources are efficiently allocated and scheduled to the clients. It follows four steps; firstly it identifies the active PMs. Next it defragments the identified machines. Then it balances the load along with the threshold feature to enhance the usage of the resource utilization. Finally it allocates the efficient Virtual Machines (VM) to the data owner as per the request. This is done us-ing cloudsim along with java.

Full text

Copyright © 2018 G Soniya Priyatharsini, N Malarvizhi. This is an open access article distributed under the Creative Commons Attribution Li-
cense, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
International Journal of Engineering &Technology, 7 (1.7) (2018) 21-26
International Journal of Engineering & Technology
Website: www.sciencepubco.com/index.php/IJET
Research paper
RHEA: resource hypervisor and efficient allocator in cloud
G Soniya Priyatharsini 1*, N Malarvizhi 2
1 Research Scholar, Department of Computer Science and Engineering, School of Computing, Vel Tech Rangarajan
Dr. Sagunthala R&D Institute of Science and Technology, Avadi, Chennai-62, TamilNadu, India
2 Professor, Department of Computer Science and Engineering, School of Computing, Vel Tech Rangarajan
Dr. Sagunthala R&D Institute of Science and Technology, Avadi, Chennai-62, TamilNadu, India
*Corresponding author E-mail: sonigeorge85@gmail.com
Abstract
In this modern world, people are not ready to waste their time in waiting for long duration. That’s why cloud computing is such an enor-
mous number of fans that it can be rented and also pay per use. The cloud service provider is concern about the data owner’s satisfaction
in cloud usage. The main area they concentrate will be the security of the owner’s data and the resource allocation as per the request.
This paper explains how the resources are efficiently allocated and scheduled to the clients. It follows four steps; firstly it identifies the
active PMs. Next it defragments the identified machines. Then it balances the load along with the threshold feature to enhance the usage
of the resource utilization. Finally it allocates the efficient Virtual Machines (VM) to the data owner as per the request. This is done us-
ing cloudsim along with java.
Keywords: Cloud Resource Management; Virtualization Hypervisor; Allocation of Resources in Cloud.
1. Introduction
Cloud is the widely used technology in the Information Technolo-
gy world. It reduces the human working power. It also reduces the
storage capacities of the computer hardware. The cloud computing
environment can be explained by a figure 1.
Fig. 1: Cloud Computing Environment.
Inspite of the tremendous growth in the cloud computing, the cli-
ents are increasing in the cloud [1].
There are many features in the cloud. Some of them are:
1) Pay as we go- The client can pay for what he/she used.
2) Interface- The cloud computing environment is the user
friendly interface.
3) Workspace- The working environment in cloud is the client
requested and customized servicing area.
4) Virtualization- This techniques reduces the cooling con-
sumption, usage of hardware etc.
The growth poses a huge threat of security and the resource provi-
sioning problems to the service provider. So to use the resources
efficiently, a resource manager should be there for the proper
management of the resources. The allocation of the resources is
done with the respect of some parameters such as availability,
performance, response time, throughput, security etc.
1.1. Resource management
Research management involves provisioning of the resources [2],
resource allocation and resource monitoring.
i) Resource provisioning: - It gives the search result for the
adequate resources types which are available and are match
the application requirements.
ii) Resource scheduling: - The procedure of assigning of avail-
able resources to the requested clients.
iii) Resource Monitoring: - It controls and also manages the
software and the hardware infrastructure of the cloud.
1.2. Virtualization
It is the process where one system which has the ability of appear-
ing likes the many systems. [3] It is used to improve the perfor-
mance parameters by using the physical resources as a pool. From
here the virtual resources can be allocated. Multiple virtual sys-
tems can run on single Physical Machine (PM).
The Virtual Machine architecture is shown in figure 2:
Applications
Applications
Applications
VM
Guest OS
VM
Guest OS
VM
Guest OS
Virtualization Platform (VMware, Xen…)
Physical Hardware
Fig. 2: Virtual Machine Architecture.
A hypervisor or a virtualization manager is a program that allows
multiple operating systems to share single hardware host. Every
Location 2
Location 1
Clients, Employees,
Corporate Users
Cloud
Service
Application
Delivery Network
Appliation
Delivery Network
Data Center
Application
Delivery
Controller
Application
Delivery
Controller
Data Center

22
International Journal of Engineering & Technology
guest operating system appears to have the host’s processor,
memory, and other resources all to itself. However the hypervisor
is actually controlling the host processor and resources, allocating
what is needed to each operating system in turn and making sure
that the guest operating systems cannot disrupt each other.
Virtualization concept is used to share the resources. It helps in the
cost reduction. This machines are isolated if this machines are
physically separated. These also encapsulate a complete compu-
ting environment. They run independently of underlying hardware.
[4] They can be migrated between hosts.
The main components of the resource management systems are as
follows:
a) Inputs: - The policies of admission control and the offered
workload, the load balancing, the energy optimization and
the QoS.
b) Components of the system: - To estimate the relevant
measures of the performance and controller, the sensors are
used.
c) Outputs-: The resource allocation to the individual applica-
tions.
The rest of the paper is organized as follows: Section 2 explains
the previous paper related to our work. In Section 3, proposed
system of RHEA is explained in detail and describes how the re-
sources are segregated for the scheduling process. In Section 4 the
process of the allocation and scheduling of the resources are ex-
plained. Section 5 explains and analyzes the simulation results of
the proposed method RHEA. Finally the conclusion of the work is
explained in Section 6
2. Literature review
Nowadays the requirements of the cloud computing users are high
and emerging widely. Thus to meet their requirements the cloud
providers are trying to do their work efficient with quality of ser-
vice. Some of the papers are discussed here on the basic of the
efficient resource management.
Mohammed et al. [5] discussed that the multi tier cloud system’s
resource allocation. It addresses the problem of over provisioning
in the enterprise class application. It proposed a multi tier cloud
provisioning for reducing resources over provisioning. T-BICA is
used to progress the process of the cloud managed resource alloca-
tion process .Also it prevents the over provisioning of the process.
It is cost analysis and the Tier centric business impact. The lever-
aging of the existing layer is introduced.
Hong Xu et al.[6] has proposed an framework for a versatile and
also an efficient resource management, Anchor. Anchor is archi-
tecture of the general resource management. It separates policies
from mechanism using the stable matching framework. This is
done while mapping VMs and physical servers. In this Anchor, a
variety of different resource management policies are able to ex-
press by the operators and the clients. The important factor about
the Anchor is the many to one. This efficiently matches the heter-
ogeneous resource needs with the VMs using online and offline
algorithms.
Sheng Di et al. [7] proposed the payment minimization and the
error tolerant allocation. The cloud systems which are in the cloud
can be clubbed. Then it can be allocated on demand. The contribu-
tions of this paper are: - It validates the effectiveness of the VM
facilitated cluster environment. The polynomial time with VM
technology is proposed based on the cloud environment. It can be
reduced the cost of allocation.
Ping et al. [8] describes about the model where the resources are
managed in hierarchical. This paper resolves the long waiting time
and the heavy load in the centralized resource management. Re-
source tables used to clear this. It gives a resource allocation pro-
cess with a model of resource management system.
Mayank Mishra et al. [9] proposed resource management which is
dynamic along with the VM migrations. The efficiency of the
resource management is increased by the VM relates features such
as isolation, flexibility, migration of the machine state. The work
explains about the usage and the process of the relocation of the
VMs. This is towards resource management dynamically in the
environment which is virtualized.
Abirami et al. [10] suggests that the response time and the waiting
time is forced in the allocation of resources. This work concen-
trates on the new method called linear scheduling of resources and
tasks. With the help of the available VMs and tasks, the schedul-
ing algorithm is designed. By using this resource utilization is
increased.
Clark et al. [11] explains the VMs which is lively migrated. It also
explains the software and the hardware separation which provides
the fault management. The running devices may have different
traffic. This can be controlled by the algorithm which is men-
tioned here is rate adaptive algorithm. Migration overhead, page
dirtied rate and the bandwidth are the factors affecting the total
migration.
Shikharesh [12] describes the match making in his paper. This
paper says that match making is the important scenario than then
was the scheduling. It is because the resources are arranged first
according to the user’s request. It is then designed from the ob-
tained resource pool. It also explains that matchmaking before
scheduling makes some uncertainties.
Nilabja et al. [13] explains the pay per use scenario. The optimiza-
tion of the system is identified here. The system behavior is con-
cerned on the basis of the performance and usage of the resources.
The cost of these can be varied basis on different factors like cost
of configuration, cost of leasing resources etc.
Soramichi et al [14] illustrates the Miyakodori algorithm for the
memory reusing technique. This technique decreases the data
amount transferring in relocation lively. When it executes the
VMs may migrates. This leads the memory image to be reused.
Thus by using this the migration time can be reduced simultane-
ously with greater optimizations algorithms.
3. Proposed system
In cloud computing there are enormous area for the scope of im-
provement. The cloud provider and the cloud user’s relation must
be like a win win policy. No one should lose from their perspec-
tive. While supplying the resources the cloud provider should not
have lack of resources also cloud user should not found that he/she
is spending more. So this can be done only by using the resource
management techniques. There are many resource management
policies are available in cloud. Here this work concentrates on the
power consumption and cost. This can be achieved by using the
following objectives.
1) Identifying the active PMs.
2) Defragmenting the identified VMs in PMs.
3) Resource Management using threshold value.
4) Efficient allocation of the cloud resources
3.1. Identifying the active PMs
The PMs which are active are recognized with the help of the
binary cuckoo search algorithm. It gives the accurate results
whether the PMs are active or not. This algorithm is already com-
pared with the other optimized algorithms such as ant colony algo-
rithm [15], bee colony algorithm [16], wolf theory [17], and parti-
cle swan optimization [18]. This yields the result as that this cuck-
oo search algorithm is comparatively give more adequate results.
The pseudo code for finding the active PMs using binary cuckoo
search algorithm is given here:
Input:Entirehostlist,Output: hostlist
leftOverVMs ← 0
for host hostlist do
for virtualmachine host.VMList do
if possible to reallocate virtualmachine in host then
Reallocate resources for virtualmachine in host
else add virtualmachine to leftOverVMs list

International Journal of Engineering & Technology
23
end if
end for
end for
for virtualmachine leftOverVMs do
Place virtualmachine in a PM in first fit manner
end for
From the above pseudocode it is clearly known that the active
PMs can be identified whether the PM is active or not. This is
because the maximum utilized servers can be arranged in the de-
creasing order. This method gives the user to view the underuti-
lized server in the last. So with respect to that of any conditions
the underutilized server is removed.
3.2. Defragmenting the identified VMs in PMs
The defragmentation of the identified PMs can be carried out with
the help of following Figures. Figure 3(a) shows the resources
without the residual fragmentation. Figure 3(b) shows the PMs
with the residual resource fragmentation.
Fig. 3 (A): Resource Allocation with More Residual Resource Fragmenta-
tion.
Fig. 3 (B): Resource Allocation with More Residual Resource Fragmenta-
tion.
The values can be derived with the help of the following formula.
Defragment=
Here,
m – Total number of active PMs
Tj – Threshold of the j-th source
Cij– Consumption of the jth resource in ith PM.
Thus the lesser is the resource fragmentation, higher the value of
the defragment. From this defragmented PMs the inactive ma-
chines can be set to sleep mode and then the active PMs can be
rearranges for the next process.
3.3. Balancing the load in the VMs
A system should have the upper and the lower threshold limit. It
can be said as overloaded, if the limit is beyond the upper limit
and can said to be underloaded, if it is said to be below the lower
limit. So by using the threshold limits, it is important for a system
to use the proper allocation techniques. This paper concentrates on
the upper and the lower threshold limits of the efficient allocation.
It is also important to minimize the migration issues. Minimized
number of migration gives more efficient results and that can be
explained with the following steps:
a) Calculate the load on the PMs and VMs
b) Calculate the lower and the upper threshold values
c) Select the host to be migrated and the PMs to accept the mi-
grated VM.
A) Calculation of Load
The load on the PMs can be found by using the summation of the
all VMs load.

=

 


Here the number of VM in mth host.
In this paper the parameters such as CPU utilization, Bandwidth
utilization and the memory utilization are considered for each
VMs. For the ith VM, the above said parameters can be calculated
as follows:

 


 


 

Then the total load on the ith vm can be calculated as follows:



B) Calculation of Upper and Lower threshold
It is observed that when the threshold value increases it reduces
the system performance. Thus it is concerned that the value of the
threshold is not much increased.
Thres=






Tupper=1-x*Thres
Tlower=0.3
There are many of the VMs are running in the PMs. So it will be
very difficult to choose the appropriate VMs for migration. If the
inadequate VM is chosen, then it raises many migration problems.
So it is important to choose the correct VM. To do the work, this
paper takes the help of the Enhanced Cloud Resource Consolidat-
ing (ECRC) algorithm for the better performance.
Pseudocode for the ECRC Algorithm is given below:
candHosts ← Ø, dispValues ← Ø
for host  PMList do
dispValues[host] ←getDispValue(host)
end for
candHosts ← getCandHosts(PMList, dispValues)
meanUtil ← findMeanUtil(candHosts)
Sort candHosts based on meanUtil
Assign a category to each candHost
for candHost candHosts do
Assign a category to each vm in candHost
Order vms in candHost in decreasing order of suitability of mi-
grating out
for orderedVm  OrderedHostVMs do
Find set of vms in candHost belong to a category in the opposite
order of suitability
if orderedVm and the set of vms can be exchanged with decrease
in resource fragmentation then
Schedule allocations of orderedVm and the set of vms
end if
end for
8
5
%
6
0
%
4
0
%
1
3
C
P
U
2
5
%
5
5
%
7
5
%
M
E
M
4
5
C
P
U
8
0
%
5
5
%
3
0
%
3
0
%
6
0
%
9
5
%
M
E
M
7
9
C
P
U
6
5
%
4
5
%
2
5
%
4
0
%
6
0
%
8
0
%
M
E
M
1
7
5
C
P
U
9
0
%
6
5
%
4
0
%
2
5
%
6
5
%
9
5
%
M
E
M
3
C
P
U
7
0
%
5
0
%
2
5
%
3
5
%
5
5
%
7
5
%
M
E
M
2
C
P
U
5
0
%
3
0
%
3
0
%
6
0
%
M
E
M
6
8
9
0
%
9
5
%
4

24
International Journal of Engineering & Technology
end for
Input: PMList, VMList, AlloSched Output: AlloSched
This gives the PMs ready for the migration list. Next the selected
VMs have to be allocated efficiently.
4. Resource hypervising and efficient alloca-
tion
The workflow of the proposed VM allocation and scheduling
method is given below:
Fig. 4: Resource Hypervisor and Efficient Allocator.
Here the filtered VMs are given as the input. Thus the scheduling
takes place with the help of the following instructions.
1) Selected VMs will be arranged in the queue on the basis of
the fitness value
2) The arrival of the VMs can be explicit on the basis of the
length of the queue.
3) Every VM is independent. It executes each one at a time.
4) When any VMs fail, then the VM waiting in the second in-
dependently do on one another.
The tasks are stored in the TQ (Task Queue). For a proper sched-
uling technique it is believed that the waiting time and the pro-
cessing time will matters. Thus this work concentrates on this area
for the better result.
Waiting time = 
 , i=1, 2, 3…n
Here R is the Residual service time and  is the system utilization.
Where,
R= 


 , i= 1, 2, 3,…n
  
 , i= 1, 2, 3…n
s- The network node’s number.
This shows the waiting process of the task. The next vision is to
process the waiting tasks with the efficient manner. i.e total time
consumed for the process of jth task. It also can be done with the
following steps.
Ti=Pij+Wi
Pij gives the time consumed by the node j to process task i.
Makespan = maxTi
Flowtime = 


Here both the makespan and the flow time should be minimized.
Thus it can be expressed as:
F=min [a x makespan + b x flowtime]
Here both a,b are weight coefficients. It should be greater than or
equal to zero and also a along with b gives 1 as output. Both are
equally important that’s why 0.5 is given for both a and b.
Procedure for the resource allocator.
Initialize the position of all the nodes and the parameters.
While iteration t is less than the maximum iterations do:
Calculate the next position of the node
Calculate the fitness value of the every candidate.
End while
Select the optimal solution and return
After completing the process work it gives the better result for the
VM placement. Thus by comparing with the existing works it
gives the better result in scheduling the VMs.
5. Experimental results
In this section, the results are expressed for various parameters. By
using the method RHEA, the performance of the resources
changed. The following Figure 5 illustrates the chart comparison
of energy consumption among the proposed system with the exist-
ing two algorithms. Those are Optimized Cloud Resource Provi-
sioning and Optimal Virtual Machine Placement. The workloads
are given in GB as inputs. The input datas are given in the Table 1.
Figure illustrates that the output received is slightly (0.009% to
0.05%) differing from the existing techniques. It is reducing much
power consumption. So it can be concluded that this technique is
useful in the energy consumption process.
i) Energy consumption
Fig. 5: Consumption of Energy Using Various Algorithms.
ii) Defragmentation of data
The following Figure 6 explains the data fragmentation of the
identified physical servers. It compares with the already existing
algorithms like OVMP and OCRP for server consolidation. The
below listed Table 2 explains the values for the data fragmentation.
In each consolidation interval, the defragmentation of the residual
resources is comparatively high for ECRC technique. So from the
experimental study it can be concluded that the proposed ECRC
technique provides better result while comparing with the existing
algorithms. The retrieved result can be gained only because of the
defragment value. Whenever the value of the defragment is higher,
then the defragment is lesser.
0
100
200
300
400
500
600
0.55 GB0.75 GB0.80 GB0.85 GB 1 GB 1.25 GB
Energy Consumption in Watts
Workloads in Gb
ECRC
OCRP
OVMP
Local control-
ler
Local
Queue
Proces-
sor
Nod
e 1
Node
m
Real time
controller
Scheduler
Adaptive con-
troller
Re-
jected
Global
US-
ER
US
ER
US
ER

International Journal of Engineering & Technology
25
Table 1: Energy Consumption Datasets
Workload in
size
ECRC
OCRP
OVMP
ECRC % on
OCRP
ECRC % on
OVMP
0.55 GB
187.2
185.4
192.3
0.009615
0.02724
0.75 GB
189.7
192.3
195
0.01371
0.02794
0.80 GB
192.1
194.3
197.4
0.01145
0.02759
0.85 GB
278.2
284.7
292.5
0.02336
0.0514
1 GB
389.4
395.2
409.4
0.01489
0.05136
1.25 GB
467.9
478.8
491.9
0.0233
0.05129
Table 2: Defragmentation Datasets
Consolidated Interval
OCRP
OVMP
ECRC
1800
64
71
65
5400
31
28
32
9000
30
24
30
12600
21
20
22
16200
26
25
28
19800
17
15
18
23400
15
10
16
27000
11
12
15
30600
13
14
15
34200
14
10
15
41400
16
12
16
45000
17
14
17
48600
17
13
18
52200
18
18
18
55800
11
10
11
59400
20
20
21
63000
30
30
32
66600
15
12
19
70200
21
18
23
73800
14
10
15
77400
14
11
15
81000
14
10
15
84600
14
11
14
Fig. 6: At Different Consolidation Intervals Residual Resource
Defragmentation.
iii) Cost
By using the ECRC technique the cost can be fairly reduced. It is
because only the active machines are in the working stage and all
the others are in the sleep mode. Thus by reducing the power con-
sumption the cost for the expenditure of power can be reduced
automatically. It also reduces the man power too. The below
shown table listed the comparison between the FOG and OCRP
algorithm with the proposed system. The numerals given here are
in dollars.
Table 3: Performance Comparision with Respect to Cost
Cost/no of tasks
FOG
ORCP
ECRC
10
3.6348
3.672
3.5229
20
6.2052
6.411
5.8303
30
11.4159
12.0167
10.6139
40
19.9727
20.522
18.7145
50
25.5633
26.7745
24.04
Fig. 4: Comparison of Existing Algorithms with ECRC.
The cost expenditure is minimized by using this technique. When
comparing with existing systems the proposed system gives 0.15 %
to 0.43% of reduction of the cost expenditure. So it can be used in
the process for minimizing the cost expenditure.
iv) Data Allocation and scheduling
To reach the convergence state, the virtual machines are allocated
with respect to the different algorithms. Initially 10 virtual ma-
chines are allocated using the FOG technique and the ECRC tech-
nique. It proves that it will be change for different techniques.
Likewise here the experiment is done for various allocations of
virtual machines. That is from 10 to 100 virtual machines. This
allocation explains the convergence state for the virtual machines.
Fig. 5: Comparison with FOG for Allocation with ECRC.
Figure 5 illustrates the allocation process for the proposed work.
This gives better result while compared with the existing works.
6. Conclusion
This paper explains a model RHEA. A Resource Hypervisor and
Efficient Allocator of the resource. This model efficiently identi-
fies the active PMs and from those machines the load is balanced
with the help of the threshold and cost as parameters. These pa-
rameters give a result in the load balancing of the resources in the
PM and also in the VM. Finally by using these parameters the
scheduling is calculated by weighing time and the processing time.
Here the flow time is minimized to achieve good results. Thus the
resources are allocated and scheduled efficiently.
0
10
20
30
40
50
60
70
80
Defragmentation data length N (bits)
Consolidated time Interval (ms)
OVMP
OCRP
ECRC
0
5
10
15
20
25
30
10 20 30 40 50
FOG
ORCP
ECRC
0
20
40
60
80
100
120
140
Number of Migrations
Consolidated Intervals (ms)
FOG
ECRC

26
International Journal of Engineering & Technology
References
[1] Puneet Himthani,” Efficient Technique for Allocation of Processing
Elements to Virtual Machines in Cloud Environment”, International
Journal of Computer Science and Network Security (IJCSNS),
VOL.16 No.8, August 2017.
[2] R.Madhumathi and R.Radha Krishnan,” Priority queue scheduling
approach for resource allocation in cloud”, Asian journal of Infor-
mation technology, 15(3):472-480, ISSN1682, 2017.
[3] Nitishchandra Vyas, Prof. Amit Chauhan,” A survey on virtual ma-
chine migration techniques in cloud computing”, Innovation in En-
gineering & Management (IJAIEM) or International journal of ap-
plication, Issue 5, Volume 5, May 2016.
[4] G. Naga Srikanth, Dr. G. Naga Satish, R. Krishnam Raju Indukuri,
Dr. P. Suresh Varma,” A novel Scheduling Model for resource al-
location in cloud computing”, International Journal of Advanced
Research in Computer Science and Software Engineering, Volume
6, Issue 7,July 2016.
[5] Mohammad firoj mithani, Shrisha rao,” Improving resource alloca-
tion in multi tier cloud systems”, IEEE International conference on
systems conference, 19-22 March 2012, pp 1-6.
[6] Hong Xu, Baochun Li, “Anchor: An versatile and efficient frame-
work for resource management in the clouds”, IEEE Transcations
on parallel and distributed systems, Vol 24,No 6, June 2013, pp
1066-1076.
[7] Sheng Di, Cho-Li Wang, “Error –Tolerant resource allocation and
payment minimization for cloud system”, IEEE Transactions on
parallel and distributed systems, Vol .24, No 6, June 2013, pp 1097-
1106.
[8] Ping Guo, Ling-ling Bu,” The Hierarchical resource management
modal based on cloud computing”, IEEE symposium on Electrical
and Electronics Engineering(EEESYM),24-27, June 2012, pp 471-
474.
[9] Mayank Mishra, Anwesha Das, Purushottam Kulkarni, Anirudha
Sahoo, “Dynamic resource management using virtual machine mi-
grations”, IEEE Communications Magazine, Vol 50, Issue 9, Sep-
tember 2012. https://doi.org/10.1109/MCOM.2012.6295709.
[10] Abirami S.P, Shalini Ramanathan” Linear scheduling strategy for
resource allocation in cloud environment”, International journal on
cloud computing and architecture, vol 2, No 1, February.
[11] Christopher Clark, Keir Fraser, Steven hand, Jacob Gorm Hanseny,
Eric July, Christian Limpach, Ian Pratt, Andrew warfield,” Live
migration on virtual machines”, 2nd Symposium on Networked
systems design and implementation (NSDI), May 2005.
[12] Shikharesh Mujumdar, “Resource management on cloud: Handling
uncertainities in parameters and policies”, CSI Communications,
edi pp 16-19.2011.
[13] Nilabja Roy, Abhisheik Dubey and Aniruddha Gokhale, “Efficient
autoscaling in the cloud using predictive models for workload fore-
casting”, Volume 3, January 2012,
[14] Soramichi Akiyama, Takahiro Hirofuchi, Ryoushi Takano, Shinichi
Honiden, ‘Miyakodori: A memory reusing mechanism for dynamic
VM consolidation” Fifth International conference on cloud compu-
ting, IEEE 2012.
[15] Visu, P., S. Koteeswaran and J. Janet, Artificial bee colony based
energy aware and energy efficient routing protocol. J. Comput. Sci.,
8(2): 227-231.2012.
[16] D.T. Pham, A. Ghanbarzadeh, E. Koç, S. Otri , S. Rahim , M. Zaidi,”
The Bees Algorithm – A Novel Tool for Complex Optimisation
Problems” Manufacturing Engineering Centre, Cardiff University,
UK, 2005.
[17] Shahla Shoghian, Maryam Kouzehgar,” A Comparison among
Wolf Pack Search and Four other Optimization Algorithms”, Inter-
national Scholarly and Scientific Research & Innovation, World
Academy of Science, Engineering and Technology, Vol: 6, 2012.
[18] Zhang Yu1 and Xiaomei Yang,” Full Glowworm Swarm Optimiza-
tion Algorithm for Whole-Set Orders Scheduling in Single Ma-
chine”, The Scientific World Journal, Volume 2013, Article ID
652061.