PNC Bank Corp. has taken a fresh approach to developing parallel processing technology. The Pittsburgh-based bank holding company intends to leverage its own systems to create a parallel-style system without making a huge investment in new hardware.
PNC's decision to fashion its own, homegrown parallel processing system to handle its consumer loans adds a new wrinkle to this burgeoning new technology, which has barely taken off within the financial industry.
The bank - with technical support from American Management Systems, a Fairfax, Va.based service company - has placed its own self-developed software on its IBM mainframe to "parallelize" the system. Now, executives there say, the bank can run its consumer loan batches in three sets concurrently, rather than in one long series, saving time and money.
Parallel processing, in its most general sense, breaks down the information or task that is being processed into a number of smaller, more workable parts that can be done concurrently, rather than in serial order. This helps banks accelerate transactions.
Commonly, the architecture for such systems consists of loosely coupled clusters of microprocessors feeding off the same pool of computer data and working simultaneously on the same task from different angles. Most of the industry's large vendors have introduced some offering for this largescale computing system to the market.
But PNC is eschewing a partnership with these hardware vendors, preferring instead to build on its existing technology.
"The system, which supports the bank's long-range strategic growth plans, processes huge volumes of data with improved speed and efficiency," said Bill Harnish, group vice president and management information systems manager for PNC's consumer credit systems. "And it offers every branch and loan center full, real-time access to all updated consumer loan information."
Although this version of parallel processing may appear too far-flung from the classical definition to qualify, various interpretations on the same theme abound. Lately, many vendors and users alike have interspersed a variety of different computing methods with parallel processing to create hybrid systems that often may bare little likeness to the traditional structure of a parallel system, but basically perform the same functions.
So, in that sense, the jury may still be out on whether or not the hardware is truly a necessary component to the parallel system. For banks, this selfstyled idea may present a cheaper and more comfortable compromise of new technology leveraging the power and stability of more familiar legacy systems.
"I think that this is a limited form of parallelism," said Richard Winter, president of Winter Corp., a consulting firm "Probably this has some practical benefit, so I wouldn't dismiss it. But it doesn't sound like the sort of system that offers major steps forward."