Designed for the application of Data-intensive computing, Data Grid focuses on data integration, data access with high performance. Data Grid Replica Location is coined to denote that get one or more physical positions by Logical File Name in Data Grid. In this paper, we provide a summary of all the possible topologies of Data Grid Replica Location Mechanisms based on the research status of related activities, and design a new kind named Distributed Replica Information Index Location Mechanism (DILM), which is based on the topology of improved ring and integrates four techniques: Overlay Network, Bloom Filter, soft-state protocol and MD5. We introduce all the components of DILM in detail, and illustrate the procedures of those components by the programming language. We also design the main APIs in DILM prototype system. The Comparison between DILM and other existing Data Grid Replica Location Mechanisms is to argue that our mechanism is of better expansibility and security, and need fewer hops of Physical Route in Replica Location. DILM can support Large-Scale Wide Area Replica Location. |