The large-scale clustering of Intel servers is one step closer to becoming reality, thanks to a massive 72-node clustering demonstration conducted by Compaq.
Recently, at Sandia National Laboratories in New Mexico, Compaq ran a trial of 72 clustered Proliant 1850R servers. The servers are part of a production network that Sandia uses for conducting government research.
The ServerNet-I-enabled cluster of Windows NT servers sorted a terabyte of data in under 50 minutes, exceeding the mark of 152 minutes previously performed by a Silicon Graphics 32-processor supercomputer.
Six to seven minutes could have been cut off the time had the Sandia team used the 10,000rpm disks presently available, rather than the 7200rpm disks Sandia has installed.
While it is unlikely that this demonstration will result in a 72-node clustering product from Compaq anytime soon, it shows that NT systems can potentially sort through huge amounts of data, like that found in decision support applications, data warehousing and Web engines.
An unreleased Aberdeen Group study shows that 25 per cent of Windows NT networks use some type of clustering or failover technology, compared with 10 per cent a year ago.
"There is an increased interest in clustering," said James Gruener, a senior analyst with Aberdeen Group. "The application environment on NT is starting to be perceived as more business-critical."
Gruener adds that as more companies rely on mail-messaging systems, they are being recognised as mission-critical applications that require high availability.