Gearing up for Grid

Financial advisers at Charles Schwab will notice a substantial improvement in their real-time retirement planning tool. That's because retirement-scenario calculations that used to take many minutes to compute are now completed in seconds thanks to a change in the way the firm has designed its computing infrastructure.

"It had been very, very difficult to get an acceptable turnaround in calculations," says David Dibble, executive vice president of technology services at Schwab. That's because it's "highly compute intensive" and the complex Monte Carlo simulations generated by the system required major computing power. It was a delay that held up advisers and left clients waiting.

So Dibble looked for a solution and decided to make the retirement tool part of his firm's first foray into grid computing, where clusters of computers are linked to tackle tasks that require heavy computing power. It allows firms to tap into their excess computing power in servers or systems that are idle or not being maximized.

Working with IBM last year, Dibble built a prototype that confirmed a grid model would speed up the task tenfold. "That speed translates into a much more satisfying customer experience," he says. Moreover, it played into Schwab's business strategy as the firm is making a push to provide advice for its customers. Part of that strategy includes using technology to make advice scalable. By making advice scalable, Schwab hopes to avoid the need to "hire an army of brokers," he says. The goal is to mix high-tech and high-touch solutions to allow advisers to better serve clients.

A team of about 15 IT staff worked with IBM for a year to build the grid system that runs the retirement tool. It comprises 12 servers at the firm's Phoenix data center, which run on Intel chips, and the system uses open-source software tools in Globus Toolkit 2.0. The xSeries 330 servers run on Red Hat Linux and IBM's DB2 database.

"The grid lets you do lots and lots of what-if scenarios," he says. But it won't stop there. Dibble says that for an electronic brokerage like Schwab, there's a lot of excess computing power that sits around. "We have to build an infrastructure to accommodate the busiest of days - two times average peak volume. That's a lot of capacity sitting around."

While the initial Schwab rollout is "fairly contained," Dibble says, "the next phase is to spread the grid beyond the machines and into the network."

Grid computing grows hot

Daniel Powers, vice president of grid computing strategy at IBM Corporation, agrees that grid computing is ripe for financial-services firms. That's because financial services is an industry that has a tremendous number of applications to run in a high-performance computing environment. Moreover, he notes, online brokers deploy a large number of servers. "When you look at the utilization characteristics of servers, it's up and down. There's a tremendous amount of unused capacity. There's a really good opportunity to mix the computing environment to fit different parts of the business."

Schwab is certainly not alone when it comes to piloting grid computing. Dushyant Shahrawat, a senior analyst with the TowerGroup in Needham, Mass., says that many firms are tire- kicking the concept, but are still in phase one of the lifecycle. Some firms are testing it in the front office, others in the back office. The real value, he says, will come when the grid is extended across the middle, back and front office. "Mass implementations are still a couple of years away," Shahrawat says, adding that no one will be there before 2006.

In the meantime, firms will approach grid computing in stages. The first phase will consist of point solutions, with firms testing grid to run certain applications within the firewall. The long-term vision of grid, however, is much grander. It includes linking all computing power within a company and even extending it beyond to a utility-computing model, where people plug into a computing grid and pay for usage. It's similar to the way that an electric grid works, where power is routed through the grid to where the demand is most required.

Risk calculations ripe for grid at RBC

However, that's the grandiose vision. For now, firms will be testing it in areas that require high-computing power. One of those areas is risk-management computation.

Don DiPalma, chief technology officer at RBC Capital Markets, part of RBC Financial Group in Toronto, is running two grid pilots. One comprises 50 servers and the other about 70 machines.

DiPalma says RBC Capital Markets deals primarily in the risk area, "making sure that the hedging exposure is right" and following the interest-rate curve so that when it changes, RBC is properly exposed. "Arbitrage requires a fair amount of computing resources," he says.

He says the firm is taking the underutilized machines in the two groups and using them to run computations. That includes tapping into excess capacity in the disaster-recovery system. He notes that post 9-11, financial-services firms have beefed up their disaster-recovery systems. It means that a "lot of machines and desktops sit around and no one touches them unless there's a disaster. We're trying to harness some of that computer power."

Now, calculations can run over night. "What used to take days to compute is now completed in several hours," DiPalma says, "What we used to do weekly, we now do nightly." As well, the firm can break down computations and do some intraday.

The faster cycle lowers the firm's risk exposure. Right now the projects are functioning independently. In the long run, he says, "We're looking to form one large grid. "What we're trying to do is put the grid in place and offer it to developers as part of the architecture. We want to push it into the software-development lifecycle, so that it becomes part of the developer's routine."

The project started more than a year ago and the bank has been working with Platform Computing of Toronto, a grid software developer that has partnerships with firms like IBM, Dell and Hewlett-Packard.

Bob Boettcher, vice president of financial services at Platform says that traditionally financial-services firms have dedicated computing resources according to each line of business. "Typically each one of these business units has purchased enough hardware to allow them to meet the peak demand," which rarely happens. He says the value proposition for grid in the risk-management area is the reduction in time it takes to make calculations. "Getting information on risk positions in five minutes might be worth millions and millions of dollars. Get the same information in an hour and it might be worthless," Boettcher says.

TradingLab adopts grid for analytics

TradingLab, an investment bank that operates in Italy, the United Kingdom, France, Germany and the United States, has also seen the benefits of grid in the risk and analytics arena.

Andrea Ceriani, the grid solution implementation project manager at TradingLab, says his firm turned to grid computing because it needed a faster way to analyze financial data.

"We found that the grid technology approach may help us in accomplishing our needs to obtain results faster ... It would allow our applications to get super-computer performance without investing in heavy-duty hardware."

TradingLab built a small test grid using technology from Sun Microsystems. The solution was implemented on a dedicated virtual Lan, using Sun's Grid Engine software and SunfireV60, with Red Hat Linux and Solaris9 X86 operating system as compute servers. The full implementation took a couple of weeks to set up. He says it is "really easy to deploy and it is one of the best solutions for cost reduction and low TCO (total cost of ownership)."

After verifying its results, TradingLab set up a production environment, which now consists of a "cluster grid" with one Sun Grid Engine server and 30 execution nodes, comprising the Sun Fire v60x Compute Grid rack system with, dual processors running Red Hat Linux.

Grid proponents say that while there are benefits to what they're adopting, it's not without challenges. Challenges from managing resources to operational issues may be difficult to overcome. However, one source from a company offering grid solutions notes that financial-services firms are willing to give it a try.

"I definitely see a lot of people thinking more strategically about what our next generation of computing platform looks like. The pipeline is very, very strong."

---

The Grid Advantage

-Running several trade scenarios in parallel, enabling more sophisticated analysis in a shorter timeframe

-Working in a heterogeneous environment and taking advantage of different operating systems and hardware

-Adding processing power quickly and easily

-Using the grid infrastructure for a variety of analytical functions

-Scheduling jobs and restarting operations more easily

-Better use of resources