When it comes to high-performance computing, the University of Toronto SciNet Consortium is king, having commissioned the construction of the fastest supercomputer in Canada. The appropriately-named SciNet High Performance Computing (HPC) project is actually comprised of two distinct, but related, ventures. One is the supercomputer itself; the other is a human component that acts as a liaison between the hardware and its users.

The supercomputer requires three megawatts of electricity—enough to simultaneously power 50,000 60-watt light bulbs, and far too much for the St. George campus’s power grid. The university will instead house the supercomputer just north of Toronto in 1,100 square meters of rented warehouse space.

IBM will be responsible for the off-campus HPC hardware—65 refrigerator-sized, front-accessible storage racks—which will comprise a union of two IBM systems. Dual-core POWER6 microprocessors will serve as the brains of the computer, while data requests and information storage will be handled by a newly designed, more efficient version of the iDataPlex server. Speed won’t be a concern, as this combination will be able to execute 360 trillion calculations per second.

Hard drive space isn’t an issue either. The supercomputer will feature a “scalable storage” system, allowing more disk space to be added to the current five petabytes without shutting down. It will be completely water cooled, eliminating the need for air conditioning and reducing energy costs by 40 per cent. The iDataPlex’s rack design makes the most of available floor space, so the SciNet HPC system will occupy a fraction of the space needed for previous iDataPlex systems, while running five times faster.

But more than just computer equipment is required to run the SciNet HPC system. That’s where the human component comes in. Optical fibre will run underground from the supercomputer to the SciNet offices located on McCaul Street. Twenty technicians and analysts will work in this facility, keeping the computer systems running smoothly. Six will function as system administrators, tending to software and hardware installations, as well as system upgrades. The on-campus site will contain a Grid Access Room, fully-equipped with cutting edge technology. From here, researchers will be able to collaborate with each other in real-time, through video conferencing.

An undertaking this substantial doesn’t come cheap. Before installation of the supercomputer, the on- and off-campus facilities required about $5.8 million in renovations. Running the on-campus facility will cost around $465,000 over the next five years, while operating the HPC centre carries a much heftier price tag (annual rent alone is $224,000). Add in costs for things like telephones, hydro, and the security system, and that number jumps to $1.85 million a year. Most of this increase is due to hydro expenses, which annually will cost an average of $1.1 million. Over the next five years, the entire project is expected to cost $50 million. Funding for the SciNet HPC venture is provided by the Canada Foundation for Innovation and the Province of Ontario via a National Platform Fund award.

Upon completion next summer, the SciNet supercomputer will be among the best in the world. It will also be relatively unique in that it will run a plethora of software, used by researchers from several disciplines, including aerospace and biomedical engineers, computational biologists, as well as planetary, theoretical chemical, and high energy experimental physicists .