next up previous contents
Next: Storage and Networking Up: High--end Facilities Previous: High--end Facilities

High-end Research Computing

High-end computing has become an indispensable research tool in most frontier fields of science, engineering and technology. Demands for computer power by researchers are driving the computer industry to deliver ever powerful computers. A fully configured Cray T3D, a massively parallel supercomputer recently introduced by Cray Research Inc., can deliver a peak performance rating of more than 300 Gflops. Many of UMBC's research projects heavily rely on high-end computers, especially for faculty hired over the last 10 years. The acquisition of an SGI Challenge XL 16-processor supercomputer, and the CRAY Y-EL, have greatly enhanced UMBC's research capabilities in the lower ranges of high--end computing.

The SGI Challenge provides a number of UMBC faculty with 2-year old state-of-the-art workstation ``supercomputing''. This machine has given a number of faculty a competitive edge that allows them to stay of the forefront of their research areas. It has an order of magnitude more memory than normal workstations, which allows research to be performed that is impossible on other machines on campus. Sharing a large, well equipped multiple-cpu computer such as the SGI Challenge is clearly the most cost efficient way for high-end computing at UMBC. In addition, management of a single high-end machine takes far fewer personnel than many smaller, less capable machines. This fact reflects the recent trend in higher-end computing away from individual machines back to shared computers. This trend is fueled by the existence of high-speed networks, that effectively make the user feel that they are using a dedicated workstation at their desk, and by the availability of inexpensive disk/memory for high end machines, that was not an option 3+ years ago.

It is essential that this capability for research computing be maintained at UMBC now that it has been integrated into a number of our most visible research programs. Our goal should be to keep our high-end computing capability at the same position, relative to the state-of-the-art, over time. Slippage in capabilities of our high-end computer, relative to the state-of-the-art, should be kept to a minimum of 2 years (maybe 3 years at most). This will require an almost continual effort to determine how to finance upgrades or new machines. With careful planning between the administration and faculty it might be possible to procure upgrades/machines at the high end with shared funding between the university and research grants.

A great amount of research computing also takes place on the roughly 200 individual departmental and faculty UNIX workstations. Because of UMBC's unique size and atmosphere of cooperation, these (UNIX) workstations have been maintained by ACS. This service takes a tremendous load off of individual researchers. Using a central organization for these tasks represents a tremendous gain in efficiency, which benefits UMBC greatly in the long run. We clearly believe that this service should continue.



next up previous contents
Next: Storage and Networking Up: High--end Facilities Previous: High--end Facilities



Dr. Larrabee Strow
Thu Mar 30 17:21:15 EST 1995