Information Technology Reference
In-Depth Information
Challenges on Amazon Cloud in Load Balancing,
Performance Testing and Upgrading
Himanshu Shah, Paresh Wankhede, and Anup Borkar
Accenture Services Private Limited,
Mumbai, India
{himanshu.shah,paresh.wankhede,anup.borkar}@accenture.com
Abstract.
Web application hosting in a data centre is clouded with quite a few
issues ranging from hardware provisioning, software installation and
maintaining the servers. Traditional data-centre techniques need production
grade hardware to test application's behavior/performance under expected peak
load. This could be costly and procuring hardware could be very time
consuming causing delays in software delivery. Cloud (Infrastructure-as-a-
Service) can be an answer to this. Cloud Computing provides production grade
server instances at very cheap rates.
This whitepaper is divided into two sub parts: first part details out the typical
web application setup on Amazon Web Services cloud (AWS) [Ref 2],
challenges faced during the setup and resolution for the same, while the second
part talks about the observations made during load testing using Apache JMeter
performance testing tool on AWS cloud. Three different application setup
topologies (single tier, two tier and three tier) are tested and findings and
learning from it are discussed here.
This whitepaper only highlights the pitfalls encountered and possible
resolutions for each and is not a comment on performance of Amazon cloud.
The whitepaper endeavors to find out the best architecture which would give
maximum return on investment.
1
Hosting a Web Application on Cloud
Organizations are lured into moving onto cloud from traditional data centre to reap
the benefits of its agility, elasticity, cost-efficiency and security.
1.1
Issues with Traditional Infrastructure
There are some issues with traditional infrastructure which increases application's
'time-to-market' parameter. Those issues could be:
a)
Hardware provisioning
b)
Installation of software
c)
Prediction of expected usage/load
d)
Scale up and Scale down
Hardware maintenance
e)
