Load balancing for Key Value Data Stores

In the last decade new scalable data stores have emerged in order to process and store the increasing amount of data that is produced every day. These data stores are inherently distributed to adapt to the increasing load and generated data. HBase is one of such data stores built after Google BigTable that stores large tables (hundreds of millions of rows) where data is stored sorted by key. A region is the unit of distribution in HBase and is a continuous range of keys in the key space. HBase lacks a mechanism to distribute the load across region servers in an automated manner. In this paper, we present a load balancer that is able to split tables into an appropriate number of regions of appropriate sizes and distribute them across servers in order to attain a balanced load across all servers. The experimental evaluation shows that the performance is improved with the proposed load balancer.