IBM Technology Fuels HPC for Financial Services

IBM provides software, hardware and services to enable financial services firms to support high-performance computing and grid systems.

NEW YORK—IBM is helping financial services customers take high-performance computing into the financial services sector.

At the High Performance on Wall Street 2006 conference here, several IBM executives highlighted IBMs strengths in the area of grid and high performance computing.

Alan King, a researcher at IBMs T.J. Watson Research Center in Hawthorne, N.Y., said at the conference on Sept. 18 that IBM provides a variety of software, hardware and services to support high-performance computing in the financial sector, including the companys WebSphere Extended Deployment offering.

WebSphere Extended Deployment manages a high volume of system transactions and simplifies the complexity of deploying applications, King said. The software has autonomic features and supports transactional grids for OLTP (online transaction processing), computational grids, and high-performance caching for object grids, King said.

Meanwhile, IBM also offers its WebSphere Front Office for Financial Markets product, King said.

This software can accept feeds directly from exchanges to help achieve high-speed delivery to machine-based trading systems and other decision-support applications, he said.

In addition, WebSphere Front Office includes a Feed Adapter Framework that lets users access a wide range of market data feeds, including direct market feeds and aggregated feeds, King said.

And the WebSphere Front Office framework and simplified API set ease feed adapter construction, maintenance, reuse and application connectivity, he added.

/zimages/3/28571.gifClick here to read more about IBMs grid computing capabilities.

Moreover, its APIs can stream data to machine, terminal and custom applications requiring streaming data.

Another area where IBM is helping financial markets customers is with its real-time Linux and Java extensions, King said.

"This is an attempt to create a real-time infrastructure for handling JVMs [Java Virtual Machines]," King said.

The real-time Java extensions also allow users to control garbage collection in real time and support ahead-of-time compiling, he said.

Ted Tso, a senior technical staff member in IBMs Linux Technology Center, said he was "excited" to be providing real-time enhancements to Red Hat Linux, "but much more interesting is what we were able to do in the Java space with real-time Java," Tso said.

Meanwhile, King said IBMs MetaCluster HPC technology, which IBM gained in its 2005 acquisition of Meiosys, is a high-performance computing infrastructure technology that allows job virtualization.

"It protects long-running compute-intensive jobs," he said.

Not only does IBM MetaCluster HPC protect long-running compute-intensive jobs, it also helps optimize usage of compute resources and application performance, King said.

Meanwhile, Steve Zuchnovich, chief technologist at Reuters, who also spoke at the conference, said Reuters is using IBM technology to empower Reuters Web services infrastructure known as Tornado.

Indeed, Reuters is using IBMs DataPower appliance technology to speed up data integration and "shape applications on the fly," Zuchnovich said.

IBMs DataPower appliances accelerate and enhance the security of XML and Web services applications and facilitate service-oriented architectures, King said.

Zuchnovich said that having software requirements migrated to a hardware environment is "where things are going. Software is hard…so you want to abstract out the software elements into hardware where you can."

Meanwhile, Kevin Gildea, an IBM distinguished engineer working in the IBM Software Strategy & Architecture group, said the IBM GPFS (General Parallel File System) "has a lot of applicability on Wall Street for scaling out applications."

GPFS avoids the bottlenecks of NFS and SAN file systems, he said.

"Your performance is scaled as the number of nodes increases," Gildea said.

"Weve been working from small scale clusters to many thousands of CPU clusters, and weve been experiencing hundreds of gigabytes per second performance."

In addition, IBM has added a grid feature to GPFS, "which allows multiple sites to group together," Gildea said.

This supports hundreds of nodes and more than 200TB of storage, he said.

Another area IBM has addressed is that of power management and cooling, King said.

An IBM software product known as PowerExecutive "gives you a visual interface where you can watch where heat is building up and address it," King said.

Kwasi Asare, IBMs System X product marketing manager, said "PowerExecutive allows you to figure out exactly whats going on with respect to your systems."

For instance, "PowerExecutive allows you to figure out per server, per rack, how much power is being consumed," he said.

Meanwhile, King said IBM also offers "deep computing capacity on demand," employing the companys BladeCenter systems and even its Blue Gene supercomputing systems.

"Well run your data center for you," King said, speaking of IBMs services group. "Well even run a Blue Gene in it for you. We have a lot of experience running secure applications and secure environments."

/zimages/3/28571.gifCheck out eWEEK.coms for the latest news, views and analysis on financial applications and services for the enterprise and small businesses.