PolyServe Inc. and IBM have introduced offerings that promise to provide enterprises with a lot of computer processing power at a relatively low cost.
A major upgrade of PolyServes Matrix Server cluster management software features a symmetric cluster file system that enables scalability with a near-linear increase in performance as more servers are added to the cluster, said company officials, in Beaverton, Ore. A distributed lock manager ensures data integrity across the cluster.
PolyServe said the upgrade, which was unveiled last week and doesnt carry a version number, is the first Linux-supported symmetric cluster file system with multiple-path I/O, which ensures that the cluster has no single point of failure. This provides greater availability to data.
The softwares support of all major Linux distributions makes it cheaper than Unix-based clusters, said Steve Norall, general manager of Linux solutions at PolyServe.
“The problem typically is you couldnt get the workload of Unix boxes” on less expensive servers with Intel Corp. chips, Norall said. “A two-way Intel server costs $5,000 today. … We let you connect lots of those boxes together.”
Matrix Server increased scalability by upping the number of nodes supported to 16. The release, available now, offers greater support for switches and broader support for Fibre Channel topologies, Norall said.
Matrix Server will add support for the iSCSI protocol at years end, Norall said. That will enable enterprises to connect their server clusters and storage area networks via Ethernet rather than Fibre Channel. This will let customers standardize on IP in Ethernet, thus reducing the number of technologies that IT departments have to manage, Norall said.
Separately, IBM, of Armonk, N.Y., late last month officially made available its Deep Computing on-demand offering—a service that lets enterprises connect to a store of clustered servers at an IBM facility via a virtual private network. The cluster, made up of Intel-based xSeries servers running Linux and pSeries AIX Unix servers, provides customers with supercomputerlike processing power without the expense of setting up and maintaining a supercomputer cluster, said Dave Turek, IBM vice president of Deep Computing.
“IT budgets are under as much scrutiny as ever,” Turek said. “People are finding [with our service] they can get a fixed amount of compute power for a fixed amount of time—rather than buying something and underutilizing it for a while.”
Buying processing power via Deep Computing on demand also relieves enterprises of the need to maintain systems and upgrade them when new hardware becomes available, Turek said.
Deep Computing on demand is targeted at uses such as for bioinformatics and digital content creation, where pattern matching and streaming information require a lot of processing power.