Fast Parallel File Replication in Data Grid

Parallel file replication where a large file needs to be simultaneously replicated to multiple sites is an integral part of dataintensive grid environment. Current data transport mechanisms such as GridFTP is mainly created for point-to-point file transfer and not for parallel point-to-multipoint transfer (required in replication). This paper presents a Fast Parallel File Replication tool (FPRF) that creates multiple distribution trees by pipelining point-to-point transfer (e.g. GridFTP sessions) and optimizes the file replication time to multiple sites. Performance results from simulations and network-level deployment of the tool in Internet show a significant speed up of up to five times using the proposed tool, as compared to using point-topoint GridFTP. 1