ACADEMIA

After obtaining impressive results in initial experiments involving eXludus Technologies' RepliCator data provisioning software, the White Rose Grid Technical Office at the University of York has signed a collaborative agreement with eXludus that includes purchasing RepliCator for the Technical Office's new cluster and demonstrating the software's capabilities on a larger scale. "The initial experiments conducted by the University of York and eXludus showed that even for a cluster with as few as 10 nodes, RepliCator greatly improved system utilization by accelerating data flow 4.4 times in comparison with the widely used NFS file system," said Aaron Turner, White Rose Grid technical manager at the University of York. "The tests were performed on our existing beowulf cluster connected with gigabit ethernet running BLAST jobs on a subset of the HomoSap dataset. We anticipate that this speed up would improve further over a larger number of nodes on the new system we will begin operating in July." Turner said the initial RepliCator test results were submitted for the United Kingdom e-science All Hands Meeting 2006, which will take place in September. The new system will include multiple dual-processor, dual-core computers connected to a cluster file system via Infiniband. The University of York's Department of Computer Science and eXludus will also conduct experiments across the White Rose Grid to determine the advantage that efficient data flow control has in a multi-site Grid. The White Rose Grid is an innovative regional computing network linking the universities of Leeds, Sheffield and York. According to Philip Bull of eXludus, the collaborators will also test the Edge version of the RepliCator software on workstation clusters and campus-level Grids. “We are extremely pleased to have made this agreement with the Technical Office of the White Rose Grid. We are particularly happy to be involved with such a leading facility within the European and global Grid community with deep knowledge of cluster and Grid environments, and we look forward to providing efficient data flow processing across compute clusters and campus Grids together,” said Philip Bull, director, European Sales, eXludus Technologies. "We expect to demonstrate that, with innovative RepliCator software, scientific and industrial research data can be transferred substantially faster within individual computer systems and between computers at different sites."