Gateway announced plans Tuesday to harness the power of its 8,000 showroom PCs that otherwise suck up resources for no particular cause.
The Poway, Calif.-based computer maker has named its foray into grid computing ‘Gateway Processing On Demand,’ a service that delivers grid computing to organizations and businesses that don’t have access to supercomputing power or the money to invest in high performance computing equipment.
The service will utilize PCs from Gateway’s 272 nationwide stores and will be based on software and network support from Texas-based United Devices, Inc., a provider of grid solutions and grid hosting.
“A lot of the decision to go in that direction was quite frankly, why not?” said Bob Moore director of New Ventures for Gateway. “We have a powerful asset that is largely unleveraged.”
According to Moore, when Gateway retail stores are open, only one percent of the time processors are being used.
“Rather than have that asset just sit there and cost a lot of money, why not use it,” said Moore. “This type of service is attractive to people who have a ton of information they need to crunch through but don’t have the time or resources to do it.”
United Devices’ MetaProcessor grid technology will provide the software framework for Gateway’s grid, along with data encryption, authentication, firewalls, and intrusion prevention methods. United Devices will also maintain the grid at an off-site data center.
United Devices’ MetaProcessor platform can produce more than 14 TFLOPs (trillion floating point operations per second) at peak capacity, and as in-store PCs are updated with the latest memory technology and Intel Pentium processors, the grid will consequently become faster, the company said.
Gateway will provide security applications at the PC level.
According to Gateway, the on-demand grid service is nearly as simple as dropping off a copy job at Kinkos and picking it up the next day. Customers submit computational job requests either via a disk or through a Web-based application that ports to a central server. A job or “problem” is then parceled out over the multicomputer grid and results are returned and tabulated to form a complete solution. Customers can also check on the progress of their project at any point during the processing period via a real-time,Web-based interface.
Sample grid computing projects include financial portfolio analysis, gene sequencing and pattern matching, oil reservoir modeling, insurance portfolio and risk simulations, and drug discovery research.
Introductory pricing is in accordance to processor hours used to complete application runs, and at present are averaging around 15 cents per processor hour. Typical grid projects average 10-20 hours, said Moore.
