Management of data intensive divisible load in cloud systems with gossip protocol

Scientific applications need to handle large volume of data-intensive computations and such computational load varies over time. We propose a new load balancing strategy for cloud environments handling those applications. Our strategy exploits the advantages of both push and pull approach to achieve load balancing. Gossip protocol is used to collect the resource statistics from a small set of cluster nodes. This paper considers a dynamic, distributed system for load balancing, which uses only the neighbourhood information to achieve the same.