Load balance optimization with replication degree customization

Abstract

The load balance optimization is an important research topic in a replication system, which will affect the performance a lot. Especially in a capacity restricted system, the load balance will be adversely affected by the strategy to replicate the object. In this paper, an algorithm which will calculate the different replication degree based on the popularity and the size of each object to improve the load balance of the system is proposed. The simulation of a cache system shows that the algorithm could achieve around 50% performance increase compare with the identical replication degree approach as the system is much more balanced and could process much more requests in a certain time period.