"We wanted to change our business model so we could deal with a customer's entire financial picture, but we were relying on systems that were built to access single accounts," Mitchell said. "If I needed to send out SOAP requests or database calls to update a single client's holdings, that could mean making 500 requests."

To speed up the process of gathering data and avoid overloading his databases, Mitchell looked into implementing a Java Message Service (JMS) system that would rely on application servers such as WebLogic. He found JMS reliable but too slow. "I needed something that could execute requests in parallel," he said.

So Mitchell turned last summer to Tangosol Inc., a Somerville, Mass. data management software vendor. Its Coherence in-memory caching software is popular with large corporations, especially financial institutions looking for faster performance. Mutual fund manager Putnam Investments last month confirmed it is deploying Coherence.

Mitchell asked Tangosol to build a distributed work manager on top of Coherence's existing features, creating a data center grid-like feature that allows data requests and retrievals to be handled quickly in parallel. They would be managed in a centralized middle tier through complex rules governing customer privacy.

"We disseminate customer information strictly on a need-to-know basis," Mitchell said, pointing out that Wells' employees have widely varying levels of access to customers' financial data. "The instinct of most technology is to make information easier to get at. We want to make sure only the right parties get the right information."