Last week, a friend, who was recently promoted within a leading software company to the role of product strategist, asked me, “When do you expect the mainframe to go away?” The scary thing is he asked it with a straight face. Even scarier were his job responsibilities, which include phasing out software products at his company for which market demand is expected to decline.
My immediate response was simple, “It will never happen.” My reasons were straightforward. First, a minimum of 30 percent of mission-critical apps continue to reside on mainframes. These wouldn’t be easily migrated to open systems, nor were they well-suited to open systems platforms from an architectural standpoint.
I supplemented this observation with tales of two recent client engagements. In one case, involving a major commodity trading organization in the Midwest, the firm was looking for a way to make a trade reconciliation database load faster. In their test environment, the lumbering application took an hour and a half to load; in their production environment, Disaster Recovery (DR) testing showed the client sitting on its collective hands for more than two hours, waiting for a start screen.
I told the client their problem wasn’t their storage or their DR strategy (the two things they engaged me to review). I asked aloud why this database, with its millions of rows and complex data paths and relationships, wasn’t running on a mainframe where it belonged. At this point, the software architect grimaced, stood up, and walked out of the room. The systems engineers smiled like the Cheshire Cat and told me once they were sure he was out of earshot, that the application had been hosted on a mainframe for many years. It was the absent fellow’s idea of strategic vision to move it onto a set of Unix boxes instead. In short, the observation I had made was exactly the one the fellow never wanted to hear, but everyone already knew.
The moral of this story to my friend at the software company: Trendiness had beaten out common sense. Don’t make the same mistake.
My other parable concerned a power company, again in the Midwest, that was working out some DR strategy issues—mostly having to do with the failover of SAP application stacks to a remote site. At lunch one day, during my visit there, I met with the CIO, a smart lady with a rather iconoclastic tendency. We discussed the challenges of her backup strategies when she flatly observed, “If I had my way, I would put all these applications back on the mainframe.”
She noted that open systems hadn’t delivered the value promised long ago by their purveyors and noted that about half her applications were only migrated (prior to her arrival) because of Y2K concerns. That, she offered, should have been a temporary thing, enabling date issues to be remediated, so the apps could return to their appropriate host. Somehow, that never happened, and now, because of the lack of mainframe-like standards in open systems (especially in storage), her staff found themselves battling the vicissitudes of SAN switch incompatibilities, unstable virtual volumes, and the myriad other problems of the Wild West of so-called open systems environments.
The moral of this story for my friend in the software industry: Mainframe discipline is more than nostalgia. In the mainframe world, best practices are designed into systems architecture and embedded in operating system controls on DASD and other peripherals. I’ve spoken to several CTO s and CIOs who are seriously looking at implementing mainframes as front-end processors to storage because the tools and techniques for storage management already exist in that space and are far better than anything available in the distributed world.
Now, my friend is an old mainframer, and he already knew these things, he said. His primary concern was who would operate the mainframes in the future. “It takes 10 years to groom a real mainframer,” he observed. Where were the educational courses and training programs to even begin the cultivation of a new generation of mainframe-savvy operators? Even IBM is having to recruit its retirees back into accounts to get bodies to support their service clients.
The shrinking labor pool is a concern, I reminded him (pointing him to my column in the previous issue of z/Journal), that’s on everyone’s mind these days—not just in mainframe data centers, but also in open systems IT. He said he was unaware that z/Journal even existed and would immediately subscribe.
So, my recommendation is for every reader to share at least one copy of z/Journal with a peer or associate to help spread the word. Heck, leave one in the reading rack in the corporate bathroom, and another in the back of the seat when you are exiting your next commercial air flight. With so much money being spent to hype the demise of the mainframe, maybe we need a grassroots campaign to reemphasize the importance of the platform and its value to the organization.