Hardware Today: On the Trail of Non-IBM Mainframes

A great many adventurers have perished pursuing their dreams. Take, for example, Robert F. Scott after years spent racing to be the first to reach the South Pole; Ferdinand Magellan en route to circumnavigating the globe; and Captain James Cook after charting much of Australasia and the Pacific islands.

Contrary to popular belief, IBM didn’t invent the mainframe. Where did the alternatives go?

Following the trail of modern-day mainframes manufactured by companies other than IBM may not be in quite the same league, adventure-wise, but it certainly poses significant problems.

In putting together this article, we hit some interesting roadblocks. NEC, for example, declined to comment on the existence of non-IBM mainframes — this from a company that announced in 2004 it was going to use Intel chips in its mainframes to cut costs. It appears further cost-cutting measures may have ended the life of the NEC mainframe altogether — at least beyond East Asia. Similarly, it took weeks of scrambling through the jungles hidden within the sprawling Hitachi organization chart to finally reveal that the company only now produces mainframes for the Japanese market.

“A few companies, such as Fujitsu-Siemens, NEC, Unisys, and Bull still sell their own brand of mainframes with proprietary operating systems,” says John Abbot, an analyst with the 451 Group. “But that’s a dwindling business.”

Few of these firms, however, are putting much promotional effort into the mainframe. The one exception is Unisys Corporation. Although its mainframe are not a corporate priority, its ClearPath Mainframes are featured right on the home page and even merit the occasional press release.

“Our customers continue to see value in our ClearPath offering,” says Steve Goldner, director of ClearPath marketing at Unisys. “So long as we continue to deliver value and new ease of use technologies, we predict a strong future for this platform.”

It’s a sign of IBM’s obvious dominance in this sector that many people consider the company to have invented the mainframe. IBM’s recent 40th birthday party for the mainframe even subtly encouraged this conviction. Yet the mainframe is a lot older, and its origins lie elsewhere.

The first commercial mainframe was called the Univac (UNIVersal Automatic Computer). It was used to process the 1952 national elections and predicted a victory for Eisenhower before the polls had closed. The Univac had already been in academic use for a couple of years before that. The company selling it, by the way, now goes under the name of Unisys.

These days, however, IBM rules and its zSeries machines are found in enterprises throughout the world. That doesn’t mean that everybody is happy about it, though. Ben Teifeld of Teifeld and Associates, a systems integrator specializing in Software AG’s ADABAS/D database, is quite critical of IBM’s influence in this arena. He feels many of the barriers to entry for new mainframe business were erected by IBM: No 110V entry-level mainframe hardware; no generally affordable emulator solution certified to run all IBM OS environments; onerous requirements for participation in mainframe development programs — hardware and software obsolescence cycles of less than one year in most cases.

“IBM will have a great deal of difficulty attracting new workloads to the mainframe because of the entry-level capital cost of the machine combined with a continuous monthly rent of software,” says Teifeld. “This is inherently more expensive for conventional Intel-based shops that are accustomed to one-time purchase and ongoing maintenance charging structures.”

One answer to IBM’s dominance came in the form of plug-compatible design — a machine designed with the same connectors and protocol interfaces to peripherals that also ran the same CPU software as an older system. The Amdahl 470 mainframe, for example, was plug-compatible with the IBM 360 and 370.

But then Amdahl was gobbled up by Fujitsu, and IBM made a strategic move that effectively killed off the plug-compatible vendors. It shifted to 64-bit, making it all but impossible for others to stay in the market.

“There are no more mainframe plug-compatibles in the old sense,” says Abbot. “The last — Fujitsu and Hitachi — pulled out when IBM upgraded the architecture of the S/390 and the compatibles couldn’t follow suit.”

According to Abbot, an Amdahl spin-off known as Platform Solutions, Inc. (PSI) is attempting to make a new pitch at the plug-compatible game. Its aim is to turn Itanium boxes into mainframes through its software. Fundamental Software is also giving it a whirl with its FLEX-ES software emulation of IBM mainframes. Teifeld says that FLEX-ES’ pricing begins at nearly $20,000 and must be run on only IBM-approved hardware configurations.

“This is the only reliably operating non-IBM solution for running mainframe software,” he says. “There are, therefore, no non-IBM mainframe solutions of real consequence, either in hardware or in software emulation.”

In such a largely uncompetitive climate, it’s refreshing to have a company like Unisys that continues to invest development dollars into the mainframe. Goldner reports three significant new features added to the Unisys mainframe in the past year: Pay-for-use allows customers to pay for computing performance as needed; application modernization lets customers leverage existing intellectual properties and assets while integrating new emerging capabilities, technologies, and open standards; and an operation and maintenance environment delivers greater automation. Further, the company is lining up a series of mainframe-related announcement for the second quarter.

“The delivery of pay-for-use and contemporary development environments, such as Java/J2EE, define a strong future for ClearPath” says Goldner. “If a customer is running business-critical operations or is concerned about TCO, the case for the mainframe is obvious.”

The new line up of ClearPath Libra servers is: Libra Model 595 with metering technology for pay-for-use; Libra Model 585, with dedicated Java processors that handle all elements of Java processing; and the Libra Model 300, an entry to midrange product. Pricing is between $50,000 to $30 million.

Old soldier may never die, but old explorers certainly do. And so do mainframe programmers, at least as viable workforce resources. Survey after survey shows an aging mainframe talent base that isn’t being replenished by an annual fresh new crop of graduates hungry to blaze new trails.

“There’s a problem looming over the next decade as more and more programmers and administrators with mainframe skills retire,” says Abbot. “It’s hard to get the new generation to learn these technologies, which they see as a dead end.”

Teifeld, however, views this situation differently. He feels that a large number of talented mainframe programmers and technicians have been hung out to dry by modern HR practices that frown on hiring workers beyond a certain age. Couple that with little or no backing from the academic community and a tendency to outsource mainframe maintenance work offshore, and its no wonder there is a shortage of trained personnel.

“There is a perceived shortage of competent professionals, largely due to discriminatory firing practices that have assured that competent mainframers over 40 cannot find work,” says Teifeld. “There is also an inherent anti-mainframe bias engendered by intellectually irresponsible university computer science and information systems curricula.”

If such problems are not confronted soon, it doesn’t really matter whether IBM has competition or not. There won’t be anyone left to oversee the mainframes that are running.