Abstract:

A new SDRL (scalable decentralized replica location method) in data grid is proposed. In SDRL replica location index nodes and replica location nodes are used to accelerate the process of locating multiple replicas of the same data file. Replica location index information are spread evenly on index nodes which can adapt to the joining or departing of index nodes. Replica location nodes support local query for local replicas. Analysis and experiments show that SDRL can locate the index nodes quickly from the logical file names and gain scalability and good load balance.

Key words: data grid； replica management, replica location； load balance