 |
HIGH-PERFORMANCE COMPUTING BUSINESS |
|
 |
|
|
|
|
HIGH-PERFORMANCE COMPUTING BUSINESS |
|
|
|
|
|
The so-called supercomputer market, dominated by major vendors selling large
proprietary mainframes (Cray, SGI, IBM, Hitachi) has long been characterized as the domain of government agencies, government-funded academic projects, and research organizations doing intense number-crunching. Beowulf clusters and Linux have expanded that range.
“It is quite possible that, by the middle of this decade, clusters in their myriad forms will be the dominant high-end computing architecture," says Cal Tech’s Dr. Thomas Sterling, developer of the first Beowulf-class PC clusters and Principal Scientist with NASA’s Jet Propulsion Laboratory High Performance Computing Group.
The dynamic in clustering’s growth in industry is Linux. Hundreds of PCs running Linux in a cluster can result in an enormously powerful machine for a fraction of the cost of a conventional supercomputer. Certainly, the Shell Oil supercomputing facility in its research center in the Netherlands using an IBM/Linux cluster (1,024 servers) has pointed the way toward clustering for high-performance computing as an approach that can fit businesses today. In our information-centric world, says an WR Hambrecht & Co. report on Linux, businesses are hungry for powerful machines, and Linux has turned the number-crunching model upside down in its use in clusters, letting today’s businesses do complex analysis on the level of supercomputing performance at PC prices. Of the TOP500 list of the world’s fastest supercomputers, Professor Hans Werner Meuer, one of the list’s co-compilers, has noted a rise in industry’s use of supercomputing. A shift is evident in the character of industry users from predominantly engineering (CAD/CAM for automotive and aerospace manufacturing) to financial services, in the latter’s investments in supercomputers for data mining rather than number crunching. What’s the alternative? Clusters and compute farms can look deceptively simple but need good advice. Experts point to a supercomputing vendor growth area: Service and consultancy firms advising on anything from Gigabit Ethernets to ventilation. |
|
|
Those Who Watch |
|
|
Back in 1993, the TOP500 project was born to track trends in high-performance (www.top500.org) computing. To achieve this, the group assembles twice yearly a list of sites operating the 500 most powerful computer systems. Recognizing the strong emergence of clusters in high-performance computing, the team that compiled the TOP500 list of global supercomputing sites is to rank the world’s top 100 cluster systems. The collection of information has begun and will continue on an ongoing basis. For now, the high-performance clusters will be ranked by peak performance only. At the same time, the team is discussing the proper choice of a benchmark for ranking clusters with the IEEE Task Force on Cluster Computing. Stay tuned and see http://clusters.top500.org/ |
|
Who’s Topping The TOP500 LIST? |
|
Installation Types |
|
Favorite
Cluster |
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|
by