Decentralized adaptive replica location mechanism in large-scale distributed systems

Efficient replica location in large-scale distributed systems is a challenging problem. In this paper, we propose a new decentralized replica location method to meet the requirement of replica location in such systems. In our method each data object has a home node, which maintains the replica location information of the data object to accelerate the process of locating multiple replicas of the same data. A dynamic balancing technique which can adapt to the joining or departing of home nodes is proposed to spread global replica location information evenly on home nodes. Soft-state strategy is used to propagate update. Analysis and simulations show that our method can achieve low latency, good scalability, reliability, adaptability and is especially fit for the environments characterized by high query rates but low rates of nodes and network changes.