After delivering the now famous ASCI White system -- the most powerful supercomputer ever created -- to the Department of Energy, IBM has commercialised the technology by putting it into a new version of its RS/6000 SP, which will be available to the public on Monday.
At 12.3 teraflops, ASCI White is the first computer to exceed the double-digit teraflop speed barrier and it earns the title of the world's fastest supercomputer with plenty of breathing room. ASCI White is 512 computers linked together to take up the equivalent of two basketball courts. It is the successor to another RS/6000 SP currently in use at the Livermore lab that can perform about 3 teraflops per second.
Rich Partridge, vice president of enterprise servers at D.H. Brown Associates, in Port Chester, N.Y. said that there is room for such a huge machine on the commercial market. He said that companies running highly technical scientific applications, risk analysis applications, and large data warehouses could have a use for it, and added that markets such as automotive, aerospace, pharmaceuticals, financials, telecommunications, and petrochemicals are prime targets to buy this system.
"Industries that have a few applications that need to really crunch numbers will find this useful," he said.
Mike Kerr, IBM's vice president of products for Web servers, says that the system is intended for companies that have workloads requiring a high degree of scalability. For instance, Kerr said that companies with Web sites that have large and unpredictable transaction rates could take advantage of the system's ability to scale by adding another node, rather than buying new systems. Kerr also pointed to companies that want to consolidate 10-20 servers into a single entity as prospective users of the new system.
"The real value is in systems management and the ability to tie all the processors together," he said.
Companies can buy a 10-processor system, for example, and allocate one processor to a firewall, one to caching, two to a database, and six to balancing application workload, thereby making those applications highly available.
"It looks and is managed as one big system," Kerr continued.
New features of the system include dynamic CPU deallocation, advanced cluster management, and LoadLeveler, which is automatic failure detection and job rescheduling software.
Additionally, the switch, software, and node of the system were upgraded. The SP Switch2 builds on the same architecture as the SP Switch, and available enhancements to the switch include hot plug capabilities, improved diagnostics, and error management.
Kerr says the software was enhanced to improve performance, reliability, availability, and serviceability.
The 375 MHz Power3 SMP High Node is available in 4- to 16-way configurations, and has four CPU card slots, four memory card slots, two disk storage bays, five PCI slots, and six I/O unit connections.
"We see customers buying in the ten to twelve node range as the meat and potatoes of the product line," IBM's Kerr said.
The basic 4-processor system costs approximately $160,000, according to Kerr.
D.H. Brown's Partridge said the commercialization of this technology was not a surprise.
"The Intent of the ASCI White program was to commercialize these kinds of technologies," he said.
(Additional reporting by Margret Johnston, a Washington correspondent for the IDG News Service, an InfoWorld affiliate.)