ISSN (Online) : 2278-1021
ISSN (Print) : 2319-5940
International Journal of Advanced Research in Computer and Communication Engineering
Vol. 4, Issue 1, January 2015
Copyright to IJARCCE DOI 10.17148/IJARCCE.2015.4125 122
Cloud Analyst: An Insight of Service Broker
Policy
Hetal V. Patel
1
, Ritesh Patel
2
Student, U & P U. Patel Department of Computer Engineering, CSPIT, CHARUSAT, Changa, Gujarat, India
Associate Professor, U & P U. Patel Dept., of Computer Engineering, CSPIT, CHARUSAT, Changa, Gujarat, India
Abstract: Cloud computing is one of the most promising computing field, which has given the new vision to the
computing field. Cloud computing has opened a door as a new model for hosting and delivering services over the
Internet. The main aim of cloud computing is to provide the resources as a services to the client. The new concept of
Federated Cloud Computing in which multiple datacenters are distributed over different regions. Since the evolution of
Cloud Computing: load balancing, energy management, VM migration, brokerage policies, cost modelling and security
issues are popular research topics in the field. Deployment of real cloud environment for testing or for commercial use
is very costly. Cloud simulators help to model various cloud applications and it is very easy to analyse. In this survey,
two cloud simulators: CloudSim and CloudAnalyst, with their overview are presented so it can be easily decided which
one is suitable for particular research topic. And also the survey on the service broker policy, its issues and available
solutions are presented. Because there is always been the requirement to select appropriate datacenter so that further
tasks for processing the request should be carried out with efficiency in least response time. So the issue of selecting
appropriate datacenter which is known as service broker policy is kind of important.
Keywords: Cloud Computing, Modeling and Simulation, Cloud Simulators, Datacenter, Service Broker Policy.
I. INTRODUCTION
Cloud computing has recently emerged as a new paradigm
for hosting and delivering services over the Internet.
Cloud computing is attractive to business owners as it
eliminates the requirement for users to plan ahead for
provisioning, and allows enterprises to start from the small
and increase resources only when there is a rise in service
demand [9].
“Cloud computing refers to computing on the Internet, as
opposed to computing on a desktop.” [1]
Nowadays applications are provided over the internet. So
there rise the need for shifting computing resources to the
service provider from the user’s location. This concept
known as “Cloud Computing”. It fulfils client’s requests
on the basis of available resources. It provides great
flexibility in terms of computing resources like storage,
platform, software, power and bandwidth. Cloud provides
better solutions to the clients of its services.
The main idea behind cloud computing is not a new one.
John McCarthy in the 1960s already envisioned that
computing facilities will be provided to the general public
like a utility [2]. The term “cloud” has also been used in
various contexts such as describing large ATM networks
in the 1990s. However, it was after Google’s CEO Eric
Schmidt used the word to describe the business model of
providing services across the Internet in 2006, that the
term really started to gain popularity [9].
Cloud computing comes into focus only when you think
about what IT always needs: a way to increase capacity or
add capabilities on the fly without investing in new
infrastructure, training new personnel, or licensing new
software. Cloud computing encompasses any subscription-
based or pay-per use service that, in real time over the
Internet, extends IT’s existing capabilities.
II. RELATED WORK
Cloud computing provides computing resources as a
service over a network. As rapid application of this
emerging technology in real world, it becomes more and
more important how to evaluate the performance and
security problems that cloud computing confronts.
Currently, modeling and simulation technology has
become a useful and powerful tool in cloud computing
research community to deal with these issues[11].
There have been many studies using simulation techniques
to investigate behavior of large scale distributed systems,
as well as tools to support such research. Some of these
simulators are GridSim, MicroGrid, GangSim, SimGrid,
CloudSim, CloudAnalyst, iCanCloud, NetworkCloudSim,
GreenCloud, MDCsim, EMUSIM, GroundSim, MR-
CloudSim, DCSim, SimIC. While the first three focus on
Grid computing systems, CloudSim is, for the best of our
knowledge, the only simulation framework for studying
Cloud computing systems [6].
CloudSim allows simulation of scenarios modelingIaaS,
because it offers basic components such as Hosts, Virtual
Machines that model the services [6].
In Section III, two cloud simulators and their main
functionalities are discussed. In Section IV, the concept of
service broker policy, its benefits, its issues and its
simulation results are presented. In Section V, available
solutions in terms of research paper are discussed. In
Section VI, implementation details is discussed. And
finally the conclusion of survey is presented.
ISSN (Online) : 2278-1021
ISSN (Print) : 2319-5940
International Journal of Advanced Research in Computer and Communication Engineering
Vol. 4, Issue 1, January 2015
Copyright to IJARCCE DOI 10.17148/IJARCCE.2015.4125 123
III. CLOUD SIMULATORS
A. CloudSim:
It is event driven simulator built upon GridSim. Its
programming language is Java, because of OOP feature
CloudSim modules can be easily extensible with user’s
requirement. It has some extra ordinary features: creating
huge datacenters, unlimited number of VMs, federated
policy, brokering policy. It supports the important feature
of Cloud Computing pay-as-you-go model. [4]
In majority of research papers,CloudSim is used forVM
Management. It consists three operations:
1. Host Overload/Underload Detection:It
determines when host is considered as being overloaded or
underloaded. It helps to decide that migration of one or
more VMs from the overloaded host. In case of
underloaded host, all VMs should be migrated and host
should be switch to sleep mode.
2. VM Selection techniques: It helps to select the
VM that should be migrated from an
overloaded/underloaded host.
3. VM Placement techniques: It helps to find out
new host for the VMs selected for migration from the
overloaded and underloaded hosts.
So there are different techniques already available for VM
Management. If someone want to focus on one of the
activity or whole VM Management for his research then
CloudSim provides best platform.
B. CloudAnalyst:
It is based upon CloudSim, adding some new features to
it. It is basically made for evaluating performance and cost
of large scale geographically distributed cloud system that
is having huge user workload based on different
parameters. It has an attractive GUI facility and flexibility
to configure any geographically distribute system such as
setting hardware parameters i.e., storage, main memory,
bandwidth etc.It gives the simulation results in terms of
chart and table that includes cost, response time,
datacenter processing time, and load over datacenter, etc.
[4]
Mainly CloudAnalyst have two separate responsibility of
VM management and service broker which is combinely
available in CloudSim. So if someone wants to focus on
particular one then he should go for CloudAnalyst.
Because it provides easy access to add new service broker
policy. It also having one extra feature that provides load
balancing among VMs that can be consider as VM
management in CloudAnalyst. [5]
In majority of research papers,CloudAnalyst is used for
load balancing policy and service broker policy.
Load Balancing Policy: It helps to select VM for
upcoming request from the user in the way that it balances
the load among VMs.
Service Broker Policy (Datacenter Selection Policy): It
helps to select datacenter. We will see this topic in brief
because we have made survey for this topic.
IV. SERVICE BROKER POLICY
A service broker decides which datacenter should provide
the service to the requests coming from each user. And
thus, service broker controls the traffic routing between
user and datacenters. So in simple words, it is datacenter
selection policy.
Fig. 1. Geographically distributed datacenters
From above figure, one can understand that in which kind
of scenario service broker policy works.Here some
datacenters and users are shown in the figure. When
request come from the user then service broker policy
helps to decide which datacenter will provide service for
upcoming request.
There are three service broker policies already existing in
CloudAnalyst [7]:
1. Closest Datacenter Policy
2. Optimize Response Time Policy
3. Dynamically reconfigurable routing with load
balancing
A. Closest Datacenter Policy:
The datacenter which is having least proximity from the
user is selected.Proximity in term of least network
latency.If more than one Datacenters having same
proximity then it will select datacenter randomly to
balance the load.
Fig. 2. Closest Datacenter Policy [7]
ISSN (Online) : 2278-1021
ISSN (Print) : 2319-5940
International Journal of Advanced Research in Computer and Communication Engineering
Vol. 4, Issue 1, January 2015
Copyright to IJARCCE DOI 10.17148/IJARCCE.2015.4125 124
B. Optimize Response Time Policy:
First it identifies the closest datacenter using previous
policybut when Closest Datacenter’s performance
(considers response time) starts degrading it estimates
current response time for each datacenter then searches for
the datacenter which having least estimated response time.
But there may be 50:50 chance for the selection of closest
and fastest datacenter. (again here random selection)
C. Dynamically reconfigurable routing with load
balancing:
This is an extension to Closest Datacenter Policy where
the routing logic is similar.But it has one more
responsibility of scaling the application deployment based
on the load it is facing. It also increases or decreases the
no. of VMs accordingly.This will be done taking under
consideration the current processing times and best
processing time ever achieved.This policy is under
research so it gives useless results.
Benefits of Service Broker Policy:
1. By using Service Broker, traffic routing between
user and datacenter is controlled.
2. It decides which datacenter should service the
requests from each user. It means it provides flexible
mapping of services to the available resources.
3. An efficient service broker policy ensures that the
later tasks to proceed the request will be done efficiently
and in least response time.
Practical comparison of these policies using
CloudAnalyst:
Now in the below table the configuration is given which is
used to take the simulation results for practical
comparison of three service broker policy.
TABLE I
CONFIGURATION DETAILS
Parameter
Value Used
UB Name
UB1
Region
2
Request Per User Per Hour
60
Data Size Per Request
100
Peak hour start(GMT)
3
Peak hour end (GMT)
9
Avg Peak Users
40000
Avg Off Peak Users
4000
DC 1 – No Of VM
75
DC 2 – No Of VM
50
DC 3 – No Of VM
25
VM Image Size
10000 MB
VM Memory
512 MB
VM Bandwidth
1000 bps
DC 1 – No Of Physical Machine
2
DC 2 – No Of Physical Machine
2
DC 3 – No Of Physical Machine
2
DC – Memory Per Machine
204800 Mb
DC – Storage Per Machine
100000000 Mb
DC – Available BW Per Machine
1000000
DC – No Of Processors Per Machine
4
DC – Processor Speed
10000 MIPS
DC – VM Policy
Time Shared
User Grouping Factor
1000
Request Grouping Factor
100
Executable Instruction Length
500
Load Balancing Policy
Throttled
In CloudAnalyst, user base configuration and VM
memory, image size, bandwidth should be define under
Main Configuration tab. Datacenter configuration which
consists no. of hosts, processor speed, memory, storage,
bandwidth, VM policy should be define under Datacenter
configuration tab. User grouping factor, request grouping
factor, instruction length, load balancing policy should be
define under Advanced tab.
Third policy gives useless results. So the comparison is
done only between first two policies. The comparison is
given below in terms of graphs for cost, response time and
datacenter processing time.
Fig. 3. Graph for cost (Comparison of policies)
Fig. 4. Graph for Response Time (Comparison of policies)
Fig. 5. Graph for DC Processing Time (Comparison of policies)
From above graphs we can say that closest datacenter
gives the best results in terms of cost and response time
compare to another two policies.
ISSN (Online) : 2278-1021
ISSN (Print) : 2319-5940
International Journal of Advanced Research in Computer and Communication Engineering
Vol. 4, Issue 1, January 2015
Copyright to IJARCCE DOI 10.17148/IJARCCE.2015.4125 125
Issues related to closest datacenter policy: [10]
Due to random selection of datacenter:
1. Possibility of selection of datacenter with higher
cost.
2. For the same configuration, results may be
differed.
3. There is also possibility of under-utilization of
the resources.
V. SOME AVAILABLE SOLUTIONS
There are some solutions available for solving the
problems arises by random selection of datacenter. There
are some solutions available in terms of research paper
and we have implemented each of them except solution 4
in CloudAnalyst.
We have classified them into two categories: Static and
Dynamic.
A. Static Approaches: [7]
Solution 1:
In this paper they have proposed that the datacenter having
less cost will be selected.Here, only VM cost is
considered. Using this solution, issue of random selection
is solved, it becomes cost effective but data processing
time is increased. For that they have proposed second
algorithm. If the two most cost effective datacenters are
selected then processing time is decreased but cost will be
little higher than the previous one.
Simulation results:
Fig. 6. Graph for Cost (Solution 1)
Fig. 7. Graph for DC Processing Time (Solution 1)
Solution 2: [10]
In this paper they have proposed that the datacenter
selection will be done in Priority based Round Robin
manner. Priority in terms of higher processing speed. It
distributes requests uniformly among all the datacenters
within a region.It leads to more resource utilization.But
cost is increased.
Simulation results:
Fig. 8. Graph for Response Time (Solution 2)
Fig. 9. Graph for DC Processing Time (Solution 2)
B. Dynamic Approaches:
Solution 3: [8]
Different datacenters may be of the same hardware
configuration but contains virtual machines in varied
number. So in this paper they have proposed that assign a
proportion weight to the data centre according to no. of
VMs it containing.According to that it handles the
resources. It makes datacenter processing time less
compare to existing ones.
Simulation results:
Fig. 10. Graph for Response Time (Solution 3)
ISSN (Online) : 2278-1021
ISSN (Print) : 2319-5940
International Journal of Advanced Research in Computer and Communication Engineering
Vol. 4, Issue 1, January 2015
Copyright to IJARCCE DOI 10.17148/IJARCCE.2015.4125 126
Fig. 11. Graph for DC Processing Time (Solution 3)
Solution 4:[3]
This algorithm selects datacenter based on two matrices
values: Cost v/s Location and Performance v/s
Availability. Here, cost means cost per VM, distance
means network latency, performance means no. of jobs
done per unit time, availability means no. of days
available for launching new VMs within year. Find
matrices values for each region and accordingly make list
of datacenters for each matrix. Take intersection of this
both list derived according to matrix value and then select
datacenter from the intersected list. It gives good result for
cost and performance both.
Fig. 12. Graph for Response time (Solution 4) [3]
TABLE II
COST VALUES FOR DIFFERENT REQUEST PER USER VALUES [3]
Request per
user
60
120
240
480
960
Closest
Datacenter
42.52
49.09
62.27
88.45
140.95
Matrix-
based DC
selection
0
0
0
0
0
VI. IMPLEMENTATION OF SERVICE BROKER POLICY IN
CLOUDANALYST
For implementing new approaches for service broker
policy, we have to make changes in following classes of
CloudAnalyst:
TABLE III
CLASSES TO BE CHANGED (FOR GUI)
Package
Class
Method
cloudsim.ext.gui.screens
ConfigureSimulation
Panel
createMainTab
cloudsim.ext
Constants
-
cloudsim.ext
Simulation
runSimulation
cloudsim.ext.servicebro
ker
Add new class
NewServiceBroker
-
TABLE IV
CLASSES TO BE CHANGED (FOR NEW APPROACH)
Package
Class
Method
cloudsim.ext.servicebrok
er
NewServic
eBroker
getAnyDataCenter
cloudsim.ext
Simulation
createDatacenter
&
runSimulation
VII. CONCLUSION
From above discussion, we can conclude that CloudSim is
limited to VM Management because in CloudSim
brokerage policy is combinely given with VM
Management which can’t be easily modified. In
CloudAnalyst these two facilities are separately given and
it also provide geographically distributed cloud
environment. So if we want to work particularly on
service broker or load balancing then CloudAnalyst is best
option. In both simulator there is no specific SLA
parameter. But CloudSim is in developing mode and they
are trying to include SLA parameter so in future it can be
there in CloudSim.
From the survey of service broker policy we can conclude
that we have to take care of two parameter cost and
performance. Till now, whichever new service broker
policy is proposed that improves either cost or
performance. So when we try to propose new service
broker policy we have to take care that both parameters
can be improved.
ACKNOWLEDGEMENT
I would like to thank Dr.RajkumarBuyya, Professor of
Computer Science and Software Engineering, and Director
of the Cloud Computing and Distributed Systems
(CLOUDS) Laboratory at the University of Melbourne,
Australia; Rodrigo N. Calheiros research fellow at
the Cloud Computing and Distributed Systems
(CLOUDS) Laboratory at the University of Melbourne,
Australia; Anton Beloglazov is a staff researcher at IBM
Research.
I would like to thank Prof.Ritesh Patel, Associate
Professor at U & P U. Patel Department of Computer
Engineering, CHARUSAT, Changa, Gujarat, India for all
his diligence, guidance, and encouragement.
REFERENCES
[1] Cary Landis and Dan Blacharski: Cloud Computing Made Easy
[2] Parkhill D (1966): The challenge of the computer utility. Addison-
Wesley, Reading
[3] AmolJaikar, Seo-Young Noh. “Cost and performance effective
data center selection system for scientific federated cloud”,
Springer Science+Business Media New York, Peer-to-Peer Netw.
Appl. DOI 10.1007/s12083-014-0261-7, May 2014