Sign In

Communications of the ACM

ACM TechNews

Supercomputing From Clusters to Clouds

View as: Print Mobile App Share:
Argonne National Laboratory Operations Director Bill Allcock

"In the future of supercomputing, [electrical] power is everything," says the Argonne Leadership Computing Facility's Operations Director Bill Allcock.

Credit: Medill Reports

The Intrepid supercomputer is the largest installation of IBM's Blue Gene architecture to date. Part of the Argonne Leadership Computing Facility, the $77 million supercomputer, the fifth fastest in the world, is used in a limited number of projects selected by the U.S. Department of Energy for their potential scientific significance. Some of the projects that will use the supercomputer this year include an experiment by University of Washington researcher David Baker designed to predict protein structures, and an Argonne project that will examine fluid flow inside nuclear reactor cores.

The supercomputer draws a massive amount of power, 1.2 megawatts, for both its computing and ancillary systems, which is why Argonne scientists are working on ways of optimizing when software runs based on generated heat. Argonne's Mike Papka says some software run hotter than others, and programs that continuously run a processor at full capacity could run approximately 20 degrees hotter than programs that run intermittently.

Part of Argonne's efforts to advance supercomputing is Nimbus, cloud computing software that allows nontechnical users to setup a virtual supercomputer at a low cost and in only a few minutes. Cloud computing can be used to pool the processing power, memory, and storage space of thousands of remote, small, and inexpensive computers to create virtual supercomputers. Nimbus is intended for scientific users, who can access a variety of science-related clouds for free, providing universities and researchers with access to computing power that previously may have been unobtainable.

View a video on Argonne's supercomputers.

From Northwestern University News Center
View Full Article


Abstracts Copyright © 2009 Information Inc., Bethesda, Maryland, USA



No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account