Grid Node Data Replication Mechanism,

In order to realize the data synchronization between grid nodes, we need to solve two important problems:replica location and replica consistency update. the paper analyzes the research of replica location recently firstly, and then we present a replica location model which is based on grid region. The entire grid system is divided into several grid regions which according with certain principles. In order to provide the function of replica location, we design the replica catalog in the paper, which is consistent of global location nodes(GLN), Domain location noeds(DLN), and storage nodes(SN). In the model, the information of replica location are stored in some distributed nodes in the grid. In order to locate the replica quickly, we search the information from near to far in the process of replica location. The replica file has a problem of inconsistency, when the update is occurred with a coincidently frequent. To solve this problem, in the paper proposed a version management for replicas consistency, each replica has a version. We also propose that each logical file is assigned a consistency update manager service(CUMS), the CUMS combined with version control mechanism can keep the consistency of replicas. In order to avoid the performance bottleneck because of the large number of replica consistency service instances, we have put those instances into some grid nodes distributed.Finally, this paper analyzes and designs the architectural structure of expert information grid, replica catalog and the replica consistency according to the idiographic requirements of expert information grid and the research results of this paper.