8:00 AM-9:00 AM
Chair: Greg Astfalk, Convex Computer/Hewlett-Packard Company
The last three years has seen a breathtaking restructuring of the HPC marketplace, driven by the victory of the microprocessor. In November, 1993, none of the Top 500 fastest installed computers in the world were from IBM, Hewlett-Packard, or Silicon Graphics. Instead almost all were from traditional vector SMP supercomputer manufacturers (Cray Research, Convex Computer) or MPP supercomputer companies (Intel Supercomputing, Thinking Machines Corporation) or their Japanese counterparts. Today, there is no stand-alone supercomputer company left and machines from SGI (having bought Cray Research), H-P (having bought Convex), and IBM (having brought out the SP-2) comprise over 50% of the Top 500 machines. Thus, we can now look forward to a market-driven battle for supremacy in the HPC market by financially stable companies with a "desktop to teraflop" product line. The Big Three (SGI, H-P, and IBM) all develop their own microprocessors, so there is also a competitive battle between MIPS, PA-RISC, and the Power Series of microprocessors. The trends we will discuss in my talk and the panel are 1) the move towards shared memory (either SMP or distributed shared memory) architectures, 2) the clustering of shared memory machines for scalability, and 3) the efficiency for the end user of writing to these machines with message passing vs. shared memory compiler paradigms. Two side issues are 1) the possible emergence of PC commodity operating systems (e.g., Windows NT) and microprocessors (e.g., Intel) as a new driving force for scalable computing and 2) whether innovative architectures and processor designs (e.g., Tera Computer) can stand against the overwhelming market forces of the Big Three.
University of Illinois, Urbana-Champaign
PP97 Homepage | Program Updates|
Registration | Hotel Information | Transportation | Program-at-a-Glance