This document discusses load rebalancing algorithms for distributed hash tables in cloud computing. It proposes a fully distributed algorithm to cope with load balancing issues. The algorithm is compared to a centralized approach and another distributed solution.
The proposed system uses multiple balancers to partition a main cloud into smaller balanced clouds or partitions. When clients upload or download files, the balancers monitor load across partitions and direct requests to less busy partitions to improve performance and quality of service. Experimental results show the distributed approach with balancers improves response time, download speed, capacity, and client satisfaction compared to no balancing.