PERFORMANCE ISSUES IN CLOUD COMPUTING

Goldi D Singh
11615327
B.Tech Student, B.Tech Department of Computer Science and Engineering
(Autonomous), Punjab, India
[email protected]
ABSTRACT
Cloud Computing makes the dream of I. INTRODUCTION
computing real as a tool and in the form
of service. This internet - based ongoing The term “cloud computing” is not a new
technology which has brought flexibility, concept for the users of computer’s world
capacity and power of processing has and the concept dates back the last decades
realized service- oriented idea and has and when John MacCarty predicted that
created a new ecosystem in the computing computers might be used one day as a public
world with its great power and benefits. utility. The confluence of technological
Cloud capabilities have been able to move advances and business development in
IT industry one step forward. Nowadays, internet broadband, web services, computer
large and famous enterprise have systems and applications has created
resorted to cloud computing and have complete storm for cloud computing during
transferred their processing and storage the past decade. Nowadays, cloud is the best
to it. Due to popularity and progress of solution for people who are looking for
cloud in different organizations, cloud rapid implementation methods. A more
performance evaluation is of special accurate and scientific definition of cloud
importance and this evaluation can help computing, models, features, performance
users make right decisions. In this paper, criteria and simulation, will be reviewed in
we provide an overall perspective on other key considerations of this paper. The
cloud evaluation criteria. For these Structure of other parts of this paper is
purposes, we present different major organized as follows: In Section ii, related
factors in cloud computing. works, simulation, evaluation methods and
tool are discussed. Section iii introduces
KEYWORDS cloud computing briefly. In section iv the
main purpose of
Cloud Computing, Performance
evaluation, Efficiency, virtualization, load
balancing.
this paper is provided. In this section, factors different users use these services. So, the
and criteria that is used in simulation will be actual cloud environments such as Google,
introduced. At the end, conclusions and Amazon EC2, GoGrid, Elastichosts, Mosso,
future works are listed. Amazon Web Service, Azure and etc. are
suitable for the evaluation, or performance
evaluation may be based on different
providers too.
II. RELATED WORK
Performance evaluation based on different
criteria, evaluation and simulation is usually III. CLOUD COMPUTING
performed based on criteria in accordance
with goals and the results are also studied Cloud computing is a type of parallel,
for proving goal. In studies related to cloud virtual, distributed, configurable, and
computing performance evaluation, some flexible systems, which refers to provision
criteria such as average waiting time, load of applications such as hardwares and
balancing and number of requests, cost and softwares in virtual data centers via internet.
throughput, workload, the rate of Cloud computing services are configurable
transactions, response time, time of and customers pay fees based on the use of
allocation and release of resources, different resources and services.
scheduling algorithms, effectiveness, delays
in service and productivity, the number of The most important element of cloud
input and output operations in the network, structure is server which is the brain behind
etc. are studied. the whole processes in cloud. Cloud is the
important model for access to distributed
Performance evaluation based on specific computing resources.
applications or services, Variety of services
and applications are offered in cloud Pay per use, scalability, use the Internet
environments each of which can be topic for technology, self-service based on the
the simulation scenarios, such as scientific demand, high performance, quick to
computing, e-learning software, high- implement, easy to maintain and update are
performance computing, inbound and key benefits of cloud computing.
outbound applications on the network, And the data recovery, lack of control over
services with assuring quality of service, cloud services, service level agreements,
Multi-Tier Cloud Applications and etc. legal problems, different architectures, audit,
Performance evaluation based on different Reviews and evaluation of the performance
environments, nowadays, large and well- cloud computing environment are the major
known organizations like Google, Oracle, disadvantages of cloud computing.
IBM and etc provide their cloud
environments with their own characteristics
and architectures, and on the other hand,
A. Cloud Computing Service Delivery Hybrid, this model is a combination of two
Models or more cloud deployment models. In this
model, resource management may be
There are three models for delivery of cloud internal or external.
services as follows:

Software as a Service (SaaS), in this model,

users use the launched application on cloud
infrastructure. Interfaces for these
applications are browsers, and don’t require IV. CLOUD COMPUTING
installation. Gmail is the best known PERFORMANCE EVALUATION
example of this model.
Cloud computing resources must be
Platform as a Service (PaaS), in this model, compatible, high performance and powerful.
users rented platforms or operating systems High performance is one of the cloud
and they can expand their required programs advantages which must be satisfactory for
on it. The most famous example of this each service.
model is Google App Engine.
Higher performance of services and
Infrastructure as a Service (IaaS), this anything related to cloud have influence on
model is associated with a virtual engine and users and service providers. Hence,
users can access to infrastructures with performance evaluation for cloud providers
virtual machine. and users is important. There are many
methods for performance prediction and
B. Cloud Computing Deployment Models
evaluation; we use the following methods in
The decision on implementation of cloud is our evaluation:
important. There are four main cloud
• Evaluation based on criteria and
deployment models as follows:
characteristics
Public, the most common model is the cloud
Another category which can be considered
deployment model. Large Enterprise is
for evaluating cloud performance is
Owner of a large cloud infrastructure and
classification of three layers of cloud
services to users.
services evaluation.
Private, this model simulates a private
A. Factors affective on performance
network. It is just for an organization's
infrastructure. Nowadays , the term “performance” is more
than a classic concept and includes more
Community, in this model, some enterprises
extensive concepts such as reliability,
which have common policies, goals and
energy efficiency, scalability and soon. Due
concerns share infrastructure of cloud.
to the extent of cloud computing
environments and the large number of
enterprises and normal users who are using optimal and more timely the
cloud environment, many factors can affect agreed requests , the higher
the performance of cloud computing and its the performance will be .This
resources. Some of the important factors view also holds true for
considered in this paper are as follows: providers.
• Network bandwidth, this
• Security, the impact of factor can be effective on
security on cloud performance and can be a
performance may seem criterion for evaluations too.
lightly strange, but the impact For example, if the
of security on network bandwidth is too low to
infrastructure has been provide service to customers,
proven. For example, DDoS performance will be low too.
attacks have wide impact on
• Storage capacity, Physical
networks performance and if
memory can also be effective
happen, it will greatly reduce
on the performance criteria.
networks performance and
This factor will be more
also be effective on response
effective in evaluating the
time too. Therefore, if this
performance of cloud
risk and any same risks
infrastructure.
threaten cloud environment,
• Buffer capacity: if servers
it will be a big concern for
cannot serve a request, it will
users and providers.
be buffered in a temporary
• Recovery, when data in cloud
memory. Therefore, buffer
face errors and failures or
capacity effect on
data are lost for any reason,
performance. If the buffer
the time required for data
capacity is low, many
retrieval and volumes of data
requests will be rejected and
which are recoverable, will
therefore performance will be
be effective on cloud
low.
performance.
• Disk capacity, can also have
• Service level agreements,
a negative or positive impact
when the user wants to use
on performance in cloud.
cloud services, an agreement
• Fault tolerance, this factor
will be signed between users
will have special effect on
and providers which
performance of cloud
describes user’s requests, the
environment.
ability of providers, fees,
• Availability, with easy access
fines etc. If we look at the
to cloud services and the
performance from personal
view, the better, more
 services are always available, pervious sections cover the factors
performance will be increase. mentioned in the previous section but some
• Number of users, if a data of the factors will be important in special
center has a lot of users and criteria:
this number is greater than
• Average response time per unit time, this
that of the rated capacity, this
criterion will cover all factors completely.
will reduce performance of
services. • Network capacity per second (Mbps)or
• Location, data centers and unit time, the most important factor
their distance from a user’s associated with this criterion is network
location are also an important bandwidth ,availability and scalability.
factor that can be effective on
performance from the users’ • The number of I / O commands per
view. second(IOPS)or unit time.

• Average waiting time per unit time

Other factors that can affect performance • Workload(requests) to be serviced per

which are as follows: second(Mbps) or a unit of time

• Usability • Throughput (Req / Sec), this criterion will

be recovered recovery, buffering capacity
• Scalability and processing power factors.
• Workload • The average time of processing(exe / sec)
• repetition or redundancy • Percentage of CPU utilization
• Processor Power • The number of requests executed per unit
time
• Latency
• The number of requests per unit time
buffer
B. Performance Evaluation Criteria
• The number of rejected requests per unit
There is a series of criteria for evaluation of time
all factors affecting performance of cloud
computing some of which will be used in V. CONCLUSION AND FUTURE
this paper. These criteria are under WORK
development. Some of these criteria have
According to prediction and evaluation of
been selected considering the importance
cloud computing performance, we can reach
and criteria in simulation. It should be
different conclusions. As an example,
mentioned that all of criteria listed in
increasing power and speed of the data
center is not always efficient, and sometimes [1] Borko Furht & Armando
it only has additional costs. So, one should Escalante,(2010) “Handbook of Cloud
not expect to increase efficiency more than Computing”, Springer,
what was required and should find standard
based on requests and user types. Due to [2] Abah Joshua & Francisca N.
increase in development of cloud Ogwueleka,( 2013) “Cloud Computing with
computing, it is predicted that storage and Related Enabling Technologies,”
computing on personal computers will be International Journal of Cloud Computing
forgotten and all of these things will be and Services Science (IJ-CLOSER), Vol.2,
transferred into the Clouds. Therefore, No.1, pp. 40~49
architecture and evaluation of the optimal [3] Sergej Poltorak, (2011) “Cloud
and efficient data centers should be Computing: Meet the Players. Performance
performed for the future of computing world Analysis of Cloud Providers”, BASEL
through suitable prediction. UNIVERSITY COMPUTER SCIENCE
According to the review and evaluation DEPARTMENT
performed in the field of performance, cloud [4]Alexandru Iosup & Simon Ostermann &
computing still has shortages in performance Nezih Yigitbasi (2010) “Performance
evaluation and special measures are required Analysis of Cloud Computing Services for
for this work. It is better to consider delay in Many-Tasks Scientific Computing”, IEEE
evaluations or implement a criterion for TPDS, MANY-TASK COMPUTING,
evaluation of service level agreement
because these agreements are the most [5] Vladimir Stantchev, (2009)
important for the users and one can present “Performance Evaluation of Cloud
more accurate evaluation in future by Computing Offerings”; Third International
specifying type of user’s request or Conference on Advanced Engineering
specifying and distinguishing all users. Computing and Applications in Sciences
IEEE

[6] Nezih Yigitbasi, (2009) “C-Meter: A

VI. REFERENCES Framework for Performance Analysis of
Computing Clouds”, IEEE/ACM
International Symposium on Cluster
Computing and the Grid.