IBM Corp. will add some grid-computing capabilities to the Enterprise edition of its WebSphere Application Server on Friday, allowing companies to squeeze more performance from disparate Web applications running on clusters of servers through better load balancing, it announced Monday.
“This is something to bring grid capabilities to commercial customers. It’s about the ability to balance Web server workloads in a more dynamic way than has ever been possible before,” said Dan Powers, IBM’s vice-president of grid computing strategy.
Grid computing is seen as a way to deliver computing power to applications as it is needed, in much the same way that the power grid delivers electricity from many sources to where it is needed.
In this case, rather than attributing fixed functions, such as serving up Web pages or handling back-office transactions, to particular machines in a cluster running WebSphere software, the software update allows each server to take on any of those tasks, depending on workload, Powers said.
This technique allows load balancing between applications, and enables systems managers to set the parameters under which servers will take on a new role, he said.
“This would benefit any customer with multiple WebSphere deployments running lots of applications, who wants to get more efficiency out of it,” he said.
The new version, 5.0.2, of WebSphere Application Server, Enterprise Edition, will also add a feature called Automatic Backup Clusters, enabling customers to designate one group of servers to act as a back-up for another cluster, without the need to write code or get involved in complex configuration, the company said in a statement.
Users can seek advice on how best to configure a cluster to balance loads using the new WebSphere Performance Advisor tool, which advises managers how to set parameters to handle different levels of network traffic, according to the statement.
Over the next year, the company hopes to improve the load-balancing technology so as to take on load balancing for other high-performance computing applications, Powers said. In that way, for example, a stockbroker’s online trading system could borrow capacity from a back-office cluster performing portfolio analysis and optimization during busy trading hours, and then take on batch processing of the portfolios after trading hours, he said.
Machines swapping processing capacity in this way needn’t necessarily be closely connected on the network. “These kinds of applications are usually data light and processing heavy. There’s not a lot of bandwidth needed,” Powers said. The company already has a number of customers in the financial industry testing such applications, he said, although for competitive reasons they did not wish to be named.
The new version will be a free upgrade for those already running the Enterprise edition. Otherwise, WebSphere Application Server costs US$30,000 per processor, or US$15,000 if upgrading from an edition other than Enterprise, Powers said. Volume discounts and enterprise license agreements could also reduce the cost, he said. The price includes one free year of maintenance and services, the company said.