Charles Schwab & Co. in December went live with a Linux-based grid computing system in an effort to speed up some of its compute-intensive investment management applications.
David Dibble, executive vice-president of technology services at Schwab, said the grid system was jointly developed with IBM Corp. and currently connects 12 two-processor servers that are based on Intel chips and located in the discount brokerage’s Phoenix data centre. Later this year, San Francisco-based Schwab plans to begin rolling out the grid technology across a thousand or more low-cost servers with spare CPU capacity that could be tapped to help boost application performance.
Dibble wouldn’t disclose the cost of the project or the throughput that Schwab has achieved on an initial retirement planning application, citing the performance levels as a competitive advantage. But he said the system lets Schwab turn around end-user requests for retirement planning data in seconds instead of days.
“We wanted to open up a new realm of high-throughput computing for Schwab’s business applications,” Dibble said. “Things that were not thinkable just a year ago are now proving economical, and we’re working at getting more of them into production.”
It took 15 internal IT staffers working with a development team from IBM about a year to build the grid system, which links IBM xSeries 330 servers running Red Hat Linux and IBM’s DB2 database. Schwab is using Globus Toolkit 2.0, open-source software that supports grid computing applications.
The system also includes IBM’s WebSphere application server software and BEA Systems Inc.’s rival WebLogic tool, Dibble said. Schwab’s IT team hopes that, in addition to boosting application performance, the grid system will help lower total cost of ownership in its tech operations. Like most large brokerages, Schwab built its server infrastructure to handle twice the computing capacity needed during peak hours on an average day, Dibble said.
“There’s a lot of capacity lying around on just average days,” he noted. “What grid computing does is enable us to go out and recapture unused capacity in a very efficient manner.”
The retirement planning application that’s running on the grid system calculates real-life portfolio scenarios based on retirement goals, risk tolerance and preferred investments. In the future, Schwab plans to add other applications designed for investment managers and to Web-enable the software for use by individual investors.
The grid system works through a “head node,” a master server that breaks up data requests into smaller jobs and sends them to systems on the grid for processing, said Willy Chiu, an IBM vice-president. The head node then reassembles the pieces of the transaction and presents the data to end users.