September 16, 2008

Sounding the death knell for mainframe computing has become such an art form that suggesting otherwise seems nearly heretical. Yet some argue that mainframes are staging a comeback within financial services.

"Some types of users and applications need the reliability, security and scalability provided by mainframes," asserts Philip Winslow, a Credit Suisse VP and software analyst. "For them, it doesn't make sense to migrate off of the mainframe from either an architectural or economic perspective."

Further, Winslow contends, positive sales statistics back him up. "Our CIO surveys of Fortune 1000 companies forecast a mid-single-digits growth in mainframe hardware spending," he says. "And some big mainframe application providers, such as BMC, ComputerAssociates and CompuWare, are even predicting double-digit growth."

True, say mainframe watchers John Phelps and Mike Chuba, both research VPs for Gartner. "Erosion in the 1990s came from the smaller shops," notes Phelps. But sales for the dominant player, IBM, stabilized a few years ago, he relates. "Currently, we're even seeing a 15 percent growth in net new MIPS [million instructions per second] shipped, which is a true indicator of mainframe power going into the marketplace."

IBM's introduction of specialty engines for running Linux and Java is a significant reason for optimism, Chuba adds. "Mainframe users can now run hundreds of Linux virtual machines on a single box," he explains. "This not only saves on hardware spending -- such as power, space and cooling -- but can dramatically reduce site license costs as well."

Plus, Big Blue's technology investments have gone beyond specialty engines, Phelps observes. "While mainframes have traditionally powered transaction- and database-intensive workloads, ... with the new System z10, IBM has advanced the mainframe's compute-intensive capabilities," he says. "Among other things, this includes a 4.4 GHz processor chip."

Such billion-dollar mainframe R&D investments signal IBM's long-term goals, Phelps maintains. "IBM isn't treating the mainframe as a cash cow that's resting on its past glory," he says. "Instead, they're actively improving and enhancing the mainframe to keep it relevant."

The Trouble With Success

With so much good news, how could naysaying continue to gain traction? Perhaps IBM's own success is a culprit, suggests Steve Josselyn, an IDC research director. "Once IBM's major competitors, Hitachi and Amdal/Fujitsu, stopped selling plug-compatible mainframes earlier this decade, IBM lengthened its development cycle," he notes. "What previously was a nine- to 12-month cycle became two years between the System z9 and the z10."

Unfortunately, the tech industry's buzz engines expect new introductions at least annually. When months pass without a product announcement, or even a sales bump, the fear, uncertainty and doubt (FUD) machines crank up. From there, it's a familiar tune: The more time FUDsters have, the more hardened negative perceptions become and, in turn, the easier it is to leverage FUD for competitive advantage.

HP's (Palo Alto, Calif.) new strategy may be a case in point. Last fall, the company hired Michael Blum, a longtime global financial services IT guru and former IBM executive. Then, in June, HP announced an incentive program dubbed "NonStop FREEdom" explicitly wooing away IBM mainframe users.

But Blum contends his new perspective simply reflects a significant computing paradigm shift. "With service-oriented architecture [SOA], we're moving toward what some call the 'Google environment,'" says HP's worldwide financial services VP. "In data centers, this goes beyond today's multiserver distributed environments to a new model with many servers delivering multiple services to multiple users. This model eliminates any critical point of failure, such as a mainframe going down."