Bank of America Describes Grid Deployment

The Global Corporate and Investment Banking (GCIB) group at Bank of America (Charlotte, N.C., $1.1 trillion in assets) has successfully deployed grid technology using software from DataSynapse (New York).

In a presentation to the Global Grid Forum in Boston, a Bank of America technologist described the bank's grid deployment, which now reaches several major business units around the globe. "One year ago, we were just Chicago and London—we added Tokyo, Charlotte and New York," said Mike Oltman, vice president, advantage risk processing development, Bank of America. "It's a huge success for the technology and the business."

Bank of America currently has over 3,000 dedicated "engines" running application code for production tasks, and an additional 3,000 machines -- mostly user workstations and development servers -- linked to the grid for nonproduction needs.

From a technical standpoint, each engine consists of a CPU blade computer running a C++ analytics package contained within a Java interface. These engines can be accessed through a standard application-program interface (API) made available to developers across the company. Behind the scenes, "broker" software handles task scheduling and resource deployment for the engines. At a higher level, "director" software handles automatic load balancing of engines between brokers. Not only can departments within a single location use the slack capacity of computing resources in other departments, but the grid extends across oceans to supply computing capacity on a global basis.

The grid will save Bank of America in the neighborhood of tens of millions of dollars in hardware expenses over three years, according to Oltman. Other benefits include simpler deployment of new hardware, higher hardware reliability and standardized development resources.

The attendant business benefits come from the ability of traders and risk managers to get more work done, and thus gain greater insights. "The more data you get, the more you understand what's happening in your market," said Oltman. "They're coming up with things they never tried before."

The grid proved its worth from the beginning. Even with only 80 engines in the Chicago office's grid, it was able to handle a "really nasty Bermudan [option]" involving 1 million calculations in 40 minutes instead of 4 hours. As the grid has bulked up, that's been reduced even further, to 20 minutes.

As the business benefits of the grid become apparent, it's ready to spread beyond the GCIB group into other areas of Bank of America. Already, the bankers in Charlotte have tapped into the grid, and it's likely that there will be more interest given the demonstrated successes. "There's a lot more work outside of GCIB … consumer banking,real estate banking and other investment banking," said Oltman. "There are a lot of places where they do computation, data grids and reporting."

"We want to keep expanding because the business will use it," noted Oltman. "We've proven it, and the business likes it."