Fast large file distribution in data grid

With more and more data created by grid computing, large file distribution is a key issue in data grid. Current file distribution in data grid is usually a style site-to-site and this mechanism is not efficient and wastes bandwidth. In this paper, we propose a novel algorithm to distribute large file efficiently in data grid. We split the file into n subfiles and transfer each subfile to a different node. And then the node which receives a subfile transfers the subfile to other nodes. The size of each subfile is different in different network environment. We must compute the size of the subfile using our simple but efficient algorithm before file splitting. To verify the performance of our algorithm, we build a simulation system and compare with Multiunicast and FastReplica. The result shows our algorithm significantly reduce the file distribution time.