The Division of IT Research Technology Services (RTS) team, led by Don DuRousseau, has collaborated with A*STAR Computational Research Center (Agency for Science, Technology and Research) in Singapore on the Long-haul InfiniCortex project. The major concept of InfiniCortex involves connecting separate InfiniBand sub-nets with different net topologies to create a single computational resource: a galaxy of supercomputers that have the ability to provide a level of concurrent supercomputing necessary for supporting Big Data and Exascale computing. Enabled by the InfiniBand extension technology supporting transcontinental distances with Obsidian’s Longbow range extenders and a high bandwidth (10 to 100Gbps) intercontinental connectivity such as Internet2, the current InfiniCortex infrastructure is comprised of computing resources residing globally at multiple locations in Singapore, Japan, Australia, USA, Canada, France and Poland.
For this project, Dr. Jaroslav Flidr from RTS has set up a four-nodes compute-cluster at GW to connect to the InfiniCortex at A*STAR in Singapore via CAAREN and Internet2. In addition, Dr. Adam Wong also from RTS has preformed a benchmarking of a real MPI parallel application: mpiBlast on the InfiniCortex infrastructure using computing resources distributed over two countries – USA and Singapore. An experiment demonstrating the use of the MPI parallel application on InfiniCortex was shown recently in the Supercomputing Conference (SC16) in November. The promising preliminary result of executing MPI parallel applications over a Long-haul InfiniBand network has illustrated the potential of the InfiniCortex platform for global supercomputing.