Chase plans to test parallel processing for card marketing.

Chase Manhattan Corp, has joined the first wave of banks to buy advanced "massively parallel" computers to mine huge data bases.

Such systems allow large numbers of microprocessors to work together, sharing pools of computer memory while operating independently.

'More Tailored Services'

After two years of consideration, the New York money-center bank decided in June to pilot-test a massively parallel processor in its credit card division to help with marketing.

"We'll be able to provide more tailored services to customers by identifying patterns of behavior in a more focused and precise way," said Jonathan Vaughan, vice president for applications in Chase's systems technology group.

Parallel processors can perform several tasks at once, operating from the same base of information.

Chase spent several million dollars on the system, which is currently being installed. Executives at the bank expect it to be fully operational by yearend.

The banking company bought 16 nodes - each comprising eight microprocessors - from AT&T Global Information Solutions, the former NCR Corp.

Catching the Wave

"The massively parallel system enables a new type of decision analysis and support that has not been available before," said Robert L. Donald, assistant vice president for financial industry marketing at the AT&T unit. "Every bank is exploring it right now."

Although Chase officials want to study parallel processing and its price-performance ratio before committing themselves further, the credit card test is indicative of a growing trend in the banking industry.

AT&T sold similar systems recently to Banc One Corp. and NationsBank Corp., Mr. Donald said.

In the middle to late 1980s, BankAmerica Corp., Citicorp, and Chemical Banking Corp. had adopted earlier generations of parallel processing hardware from Teradata Corp., which NCR later acquired, he added.

Some banks have been reluctant to come around to parallel processing until it advanced enough to prove truly viable for their applications.

AT&T had courted Chase as a potential parallel processing customer since June 1992, Mr. Donald said.

Even now, according to Mr. Vaughan of Chase, "the technology itself is fairly immature, and there are still a lot of issues to be examined."

A Developing Technology

But the general, steady progress of technology has upgraded parallel processors to "rock-solid systems," according to Mr. Donald.

Similar to these massively parallel systems are symmetrical multiprocessing units - made by NCR and used by banks such as Chemical and Wachovia Corp.

These systems also share one pool of memory but are not linked as closely as in parallel processing. And they work efficiently in configurations only up to eight processors.

Mr. Donald added that parallel processing has become more attractive to banks because they operate within open systems.

Indeed, open systems architecture was important to Chase officials deciding whether to buy a parallel processor because the bank has also become very involved in the highly distributed processing approach known as client-server.

"These systems are now the most mission-critical systems banks will have," Mr. Donald said. "The ability to look at that detailed data is relationship banking."

For reprint and licensing requests for this article, click here.
MORE FROM AMERICAN BANKER