GOVERNMENT

Wilfrid Laurier University (WLU), one of 14 Ontario universities and colleges on the growing Shared Hierarchical Academic Research Computing Network (SHARCNET) grid, has chosen technology from Silicon Graphics for computational sciences, including mathematical modeling, nanotechnology, biology, quantum physics, chemistry, and a wide variety of scientific research projects. Installed in October, the SGI Altix system is shared with other SHARCNET members throughout southwestern Ontario and represents the first time the powerful shared memory processing of SGI architecture, which can make a huge amount of RAM available as needed, will be available on the grid. The University purchased the SGI Altix computer, with 256GB memory and 128 Intel Itanium 2 processors-one of the fastest computers in Canada-through grants from Canada Foundation for Innovation, and the Ontario Research Fund. Wilfrid Laurier University previously hosted a small 8-processor code development and testing cluster on SHARCNET. With the acquisition of the SGI Altix system WLU joins the ranks of the major grid members at the Universities of Western Ontario, Guelph, McMaster and Waterloo, who host large production clusters to accelerate research results. The SHARCNET consortium of leading academic institutions also includes the Universities of Windsor, Brock, Trent, Lakehead, Laurentian and York, the University of Ontario Institute of Technology, and Fanshawe and Sheridan Colleges. Production clusters are located in individual sites but they are equally available to all of SHARCNET. Ultimately, the goal is that researchers will just submit jobs without knowing which machine their job will run on, or where that machine might be located. Jobs with large datasets that need shared-memory architecture and very powerful processing would automatically be routed to WLU's SGI Altix system. "A lot of work is planned for the new SGI Altix system because many researchers throughout SHARCNET have a special need for big memory jobs," said Dr. Ian Hamilton, professor of theoretical chemistry, Wilfrid Laurier University, and the Laurier lead on this acquisition. "Many problems can be broken down and run serially, but some critically important problems can't be. Whenever you have a system with many interacting components, there is a real advantage to using the SGI shared-memory architecture because you can look at all aspects of the problem. Instead of running pieces of molecules and portions of datasets on machines less powerful than the SGI Altix, researchers will now be able to look at the entire problem." In nanoscience, calculating properties of quantum dots, rods or wires involves modeling an assembly of atoms that are interacting with one another, leading to the overall behavior of interest. There are many other examples. In biology, a researcher looking at the genome wants to be able to have the entire structure in memory, to be able to examine different aspects of it simultaneously. In chemistry and biochemistry, large molecules often need to be simulated. A researcher looking at photosynthesis wants to be able to have the entire structure in memory because something which happens in one region of the molecule can very quickly affect another region. When Laurier decided to upgrade their on-site computational capacity and enhance their contribution to SHARCNET, they determined a definite need for a large, symmetric multi-processor (SMP) machine. A request for proposal (RFP) was issued without specifying the number of processors beyond the minimum of 64 that the committee, which included Dr. Hamilton, had decided upon. "Wilfrid Laurier is the only SHARCNET site with shared-memory architecture, so that was a very important component of the infrastructure. Linux was something that we insisted on because there's no propriety code and porting Fortran or C is actually very easy. We're really delighted to have gone with Silicon Graphics; the Altix will get a lot of use," added Dr. Hamilton. The SGI Altix 3700 system, installed in October, is linked to desktops in research areas at WLU as well as being linked via 10GigE to the four main SHARCNET sites, and by 1Gb fiber optics to the other SHARCNET partners. "We expect the SGI Altix system at Wilfrid Laurier University will be in great demand by the entire SHARCNET community, which is already involved in simulating the formation of cosmic structure, modeling the movement of urban pollution, investigating immune systems and drug resistance, and developing stronger lightweight alloys," said Michael Brown, application segment manager, scientific research at SGI. "SGI Altix systems are excellent grid resources since they allow users anywhere on the grid to quickly and easily generate unique results that are only possible with a combination of large shared-memory and scalable computing power."