NETWORKS

HP today announced that it will supply Canada's Shared Hierarchical Academic Research Computing Network (SHARCNET) with four new high-performance computing clusters that will increase the network's capacity from 1,000 to 6,000 processors. The $20 million upgrade will enable SHARCNET, which includes 11 leading Ontario academic institutions, to meet the rapidly growing high-performance computing (HPC) demands and advance research progress for a diverse Canadian research community, and help realize the vision of establishing a world-leading computational grid. The cluster expansion is expected to assist SHARCNET in accelerating breakthroughs in such areas as human genomics, the containment of infectious human and animal diseases, improving weather prediction, simulating the collapse and formation of planets, and the development of nano-scale electronic devices. Once fully installed in early Fall 2005, it is expected that the HP Cluster Platform systems, to be housed at SHARCNET academic institutions, will combine to be the most powerful supercomputer in Canada and achieve ranking on the Top500 supercomputers list. In addition, SHARCNET, a recognized pioneer in grid computing, will be a test bed for innovative, scalable HPC solutions in data storage and visualization over a distributed network. The 500-terabyte data storage facilities will be equivalent to tens of thousands of today's top-of-the-line personal computers. The visualization systems will allow researchers to visualize enormous data sets in fields ranging from bioinformatics to astrophysics. "The HP Unified Cluster Portfolio provides us with an integrated, open-source based supercomputing utility to support distributed, high-performance computing, data management and visualization," said Hugh Couchman, scientific director, SHARCNET. "Our researchers will be able to use as much, or as little, of the processing and data capacity as they require, and seamlessly access it from anywhere on the network. Such technology is critical to Canada's competitiveness as a world-class research community and our ability to accelerate the production of results that are of benefit to our economy, health, environment, scientific knowledge and culture." The new HP clusters, which are collectively expected to deliver more than 25 teraflops of performance and total 1,900 nodes, consist of HP Cluster Platform 4000 systems based on HP ProLiant servers with AMD Opteron processors running HP's Linux-based XC System Software. With a full Linux operating system, cluster management functionality and scalability enhancements, robust resource management and scheduling with Platform's LSF, and HP-MPI, the XC System Software provides a single, supported environment to manage a diverse and distributed user community and workload across clusters having distinct design attributes. Cluster interconnection will be delivered by GbE, Quadrics ELAN4, Myrinet and Voltaire's InfiniBand. Of critical importance to many HPC scientific and engineering applications, such as those running at SHARCNET, is the availability of storage and visualization systems that can scale with increases in computing capacity. A balance of all three capabilities -- computation, storage and visualization -- is required. SHARCNET's upgrade will integrate a 500-terabyte HP StorageWorks Scalable File Share (HP SFS) system to provide a single, sharable, very high-bandwidth file system to each cluster. In addition, an HP scalable visualization system will deliver shared, high-performance, low-cost visualization across the clusters. The HP Cluster Platform, XC System Software, HP SFS and scalable visualization system are part of HP's Unified Cluster Portfolio announced in November 2004. "The ability to feed the compute engine with large volumes of data is increasingly critical for tackling more complex scientific and engineering problems, and the ability to see that data visually is key to allowing researchers to absorb information more quickly," said Winston Prather, vice president and general manager, High Performance Computing Division, HP. "We are extremely excited to help SHARCNET move to the forefront of HPC and advance collaborative research with high-performance clusters." Established in January 2001, SHARCNET is structured as a "cluster of clusters" across south-central Ontario to meet the computational needs of researchers and to facilitate the development of leading-edge tools for high-performance computing grids. The HPC infrastructure expansion is enabling four additional Canadian universities to join the research community -- the universities of Waterloo, Brock, York and Ontario Institute of Technology. SHARCNET now includes 11 leading academic institutions, the other members are: the universities of Western Ontario (lead institution), Guelph, McMaster, Wilfrid Laurier and Windsor, and Fanshawe and Sheridan colleges.