Mumbai, July 09: Bhabha Atomic Research Centre (BARC) has developed a very high speed ANUPAM-XENON/128 supercomputer, achieving another significant milestone in the field of supercomputers. The computing speed of this 128 processor ANUPAM supercomputer is 202 Giga Floating Point Operations Per Second (GFLOPS) on High Performance Linpack benchmark program and is about three times faster than the 64-node supercomputer developed in July 2002, Head (computer division) H K Kaura said on Wednesday.

This 128 ANUPAM supercomputer is built using 64 dual xenon servers as `compute nodes' in a cluster, interconnected by a high speed communication network. Each server is based on dual xenon, 2.4 GHz processors, with 2 GB memory, and 40 GB hard disk.
The present supercomputer is also more than 6,000 times faster than BARC's first 4-node supercomputer developed in December 1991. So far BARC has developed 16 different models of the ANUPAM series of parallel supercomputers using a variety of processors as `compute nodes' and various technologies for interconnection networks, Kaura said. The inter-communication network is designed using Scalable Coherent Interface (SCI), with a very high node-to-node communication speed of 300 megabytes per second and a very low latency of 3.5 microseconds, the scientist added.
The open source Linux operating system is used on each parallel processing node. For implementing parallel processing applications on the ANUPAM system standard Message Passing Interface (MPI) is available.
Bureau Report