Load Balancing Using Distributed Cluster

Access to high performance service on the web is a major challenge faced by web service providers. Performance of web service could be improved with high end configured fast servers. This could be an ideal situation if the web service is accessed consistently throughout the year. But these high end servers would be a burden on organization providing web services with variations in the web traffic pattern. These organizations are looking for alternatives to provide fast web access to their respective traffic patterns. This could be done by distributing the load across a cluster configured with only commodity hardware instead of high end expensive servers. We present a proof of concept for the above using single server performance measured using jmeter and distributed cluster using hadoop and testdfs I/O provided by the apache foundation. We are comparing the access speed of files on a server with access speed of files on distributed cluster. We are taking throughput as the criteria to assess individual system performance.