Oracle rejects grid computing assault

Oracle has emphatically rejected claims by a leading figure in the European grid research community that vendor grid computing offerings were "overhyped".

Oracle has emphatically rejected claims by a leading figure in the European grid research community that vendor grid computing offerings were "overhyped".

Roland Slee, Oracle Asia Pacific's senior director, business and technology solutions, said grid computing -- where computing power and data storage capacity can be shared between computers over a private network or the Internet -- is "not being overhyped," but rather a new architecture that delivered greater performance and flexibility at reduced cost.

Slee was speaking to ZDNet Australia   in response to remarks made by Mark Parsons, commercial director of the Edinburgh Parallel Computing Centre, who last week said grid computing was still far from achieving its full potential.

"It's clear that grid has a lot to offer, but it's still early days," Parsons added. "There's been far too much hype about it".

Parsons' comments have struck a raw nerve as vendors try to overcome customer confusion over what precisely grid computing is and wariness over sharing of hardware resources and data with other groups. He pointed the finger at the number of words used by vendors to describe their grid computing strategies. "Autonomic computing, adaptable computing, cluster computing, on-demand computing, utility computing, agile IT -- all of us see these terms, which are different names for the same thing," he said. "One of the problems in this area is that there are far too many marketing departments involved."

However, Oracle's Slee cited the success of Project MegaGrid -- a standardised enterprise grid project launched late last year in the United States by heavyweights EMC, Dell and Intel as well as Oracle -- as an example of actual grid computing at work.

He said the partners had successfully tested the performance and cost aspects of the grid using a telecommunications network provisioning application called Cramer 4.

The application had originally been benchmarked on a 72-CPU Unix server in a configuration that produced 550,000 transactions per hour with sub-second average response times.

"I understand this single server had a list price in excess of US$2 million," the Oracle executive said.

"The MegaGrid partners have demonstrated that Oracle database 10G running on a grid of just 10 Dell 1750 servers running Red Hat Linux and EMC storage can provide this same level of performance, but on servers that cost just US$60,000.

"That is a more than 90 percent reduction in the cost of the server infrastructure needed to support this real-world application," he added.

Slee also took exception to Parsons' criticism that Oracle's grid computing products of Oracle were "merely a new spin on clustering".

"Grid computing is not just a spin on clustering, rather it is an architectural evolution beyond clustering. Customers used to run business applications on dedicated servers; that's called scale-up computing.

"Then they were able to run applications on groups of servers, but with each application having its own dedicated group; that's called clustered computing. Now customers are able to run multiple applications on a pool of servers, dynamically provisioning resources from that pool as their applications demand them; that's called grid computing," Slee said.

ZDNet UK's Ingrid Marson contributed to this report.