A Prefetching-based Replication Algorithm in Data Grid

Replica value determination plays an important role in the problem of replication in Data Grid, since only replicating high value replicas can improve the replication efficiency. However, replica value determination methods applied now do not perform well, because they cannot well adapt to the changes of file access patterns. We address this problem by introducing a new replica value determination method through data mining, and based on this method we propose a prefetching-based replication algorithm in a virtual organization (VO) environment in order to carry out a better replication optimization. The major idea of this algorithm is to make use of the characteristic that members in a VO have similar interests of file. Experiments show that compared with traditional replication algorithms our new replication algorithm shows better performance and efficiency in Data Grid.