ANUPAM - Ameya: A teraflop class supercomputer

Abstract

High Performance Computers are essential tools of trade in modern day science. The current research problems scientists and engineers are attempting to solve are becoming increasingly complex and are beyond the reach of traditional laboratory style experimental methodologies. Access to supercomputers makes a great deal of difference as they help in dealing with very high resolution models which result in accurate simulations of the phenomena that one is trying to investigate, in reasonable timeframes. This, in turn helps in cutting down the number of experiments that one has to conduct and also in arriving at accurate inferences at a faster rate. The Computer Division of the Bhabha Atomic Research Centre (BARC) has been developing high performance parallel computers for its scientists and engineers for over a decade. The latest in the ANUPAM series of parallel supercomputers is called 'Anupam-Ameya', a 512-processor cluster running Linux. This paper describes the design goals, architecture, components and benchmarking figures of the Anupam-Ameya cluster. © 2007 Springer.