ATM maximizes Supercomputer at Baylor College of Medicine

Houston, 17 October 97 In April of 1996, the Core Visualisation Laboratory at Baylor's Medical School unexpectedly received a Cray Origin2000 supercomputer as part of a grant. Chris Young, the College's director of network services, had been occupied during a year with the building of an FDDI network to secure a two-and-a-half-mile Ethernet backbone when the machine turned up. At that moment, Young decided an Asynchronous Transfer Mode (ATM) network would ideally meet with the high speed performance qualities of the supercomputer.

Baylor's researchers consider the Cray Origin2000 as a critical addition offering high-speed visualisation of complex images, such as individual strands of HIV DNA, and enabling the use of a virtual reality helmet and gloves. The institute can be placed among the top twenty medical schools for research funding but was suffering from several weekly brownouts and monthly blackouts due to the insufficient Ethernet functioning.

According to Young, the supercomputer simply outperforms the FDDI network but now he needs to provide e-mail and word processing as well as high-quality transmission for virtual reality tasks. ATM technology seemed to offer a more than satisfactory solution because of its built-in scalability and therefore Young conducted some preliminary testing and research during the summer and fall of 1996.

Another factor which made Young opt for ATM was speed. He evaluated systems from Cisco, Fore and Newbridge Networks and ran each vendor's products on the other vendor's proposed architecture to find out how routers would best be deployed to meet the new processes demanded by the supercomputer. Finally, Young selected an architecture of switched ATM routers from Newbridge, based on the Multiprotocol Over ATM (MPOA) standard. The network was installed from October 1996 until January 1997.

Young believes an MPOA network especially asks for quality service at the application level and for virtually sharing the intelligence of each desktop computer. On the first of January 1997, the supercomputer and ninety desktops were taken off of the legacy network and put on the ATM network. In mid-September, four hundred more desktops were cut over. The total implementation will be completed in March 1998 with final commodity applications added by September 1998.

Baylor faces a bright future with no need for significant upgrading for the next ten years. A relieving thought since the complete network will have cost about $6.5 million. Nevertheless, Young proudly testifies the project has already paid for itself with two grants won because of the supercomputer and, not to forget, broad smiling research people.

Leslie Versweyveld