Researchers at the Shared Hierarchical Academic Research Computing Network (SHARCNET) and the High Performance Computing Virtual Laboratory (HPCVL) plan to share equipment and funding in hopes this will allow them to complete projects more quickly.
HPCVL is comprised of two Kingston, Ont.-based
schools (Queen’s University and The Royal Military College of Canada), the University of Ottawa and Carleton University, also based in Ottawa.
Ken Edgecombe, executive director of HPCVL at Queen’s University, said his organization and SHARCNET want to share resources. He added the government needs assurance “”that the money they’re spending is not going to be spent on equipment that’s being duplicated ….””
At this point, both organizations have no studies forecasting the expected cost and time savings generated by joining forces. But last year, HPCVL gained access to the library of Numerical Algorithms Group, saving HPCVL researchers “”money and development time because these are all tried-and-true, proven algorithms,”” said Edgecombe.
Major projects include research into bioinformatics, genomic and proteomics, said Carmen Gicante, executive director of SHARCNET, headquartered at the University of Western Ontario in London and made up of a consortium of 11 high-performance computing clusters across south-central Ontario.
All major health sciences sites will be connected by a dedicated fibre network that allows staff to work more easily on the many projects being undertaken in Ontario institutions, said Gicante. He said this will boost security and eliminate slow-moving Internet traffic.
Eventually, researchers will find that HPCVL and SHARCNET offer different advantages, explained Edgecombe. For instance, those who do several types of calculations that best run in a distributed fashion with little communications between processes will look to SHARCNET, which has a collection of small servers.
He added HPCVL, which has 408 central processing units, 11 large servers and a large volume of memory, is suited for calculations that are memory-intensive or need fast communication between processes.