Aug 1 ’05

Don’t Outsource Your Thinking: Mainframe Viability in 10 years?

by Editor in z/Journal

Where will the mainframe be in 10 years? If z/OS is your lifeline, your risk assessment should adequately consider this point. Since one of my clients estimates it would take them eight years to migrate from zSeries to another platform, they recently engaged me to analyze mainframe viability through 2015.

Mainframe viability has been questioned for years because certain things have been shrinking, such as the number of customers, people with mainframe skills, and number of ISVs. But a viability analysis must contrast this with strong vital signs such as the technology advantages over alleged alternatives and growing customer dependencies on z/OS. The biggest threat would be if IBM decided to pull the plug on the mainframe, but that’s extremely unlikely for a product that’s so well-positioned for the future. Let me explain.

Today we see signs that the utility computing movement is gaining traction. The primary goal for a utility is for users and application developers to access whatever info they need and get computing capacity wherever they need it without having to know the physical location of the resources involved. Since an IT utility is where heterogeneous resources will be organized and managed as pools of resources, then application longevity through infrastructure “rust out” will no longer be a concern because virtualization also allows applications to be de-coupled from the physical limitations that once threatened their longevity. Infrastructure simplification through increased consolidation and centralization of scattered servers is mandatory to minimize cost. Clusters, blades, and multi-processor servers are the hardware benefactors of these trends. Those who are weary and wary of self-destructing, virus-infested PCs will eagerly embrace an affordable and service-oriented computing utility that’s self-optimizing, self-healing, self-protecting, and self-configuring.

Over the next 10 years, the mainframe, specifically the new System z9-109 family, will continue to leverage its extensive virtualization architecture combined with a number of very sophisticated hardware optimization techniques to provide significant economies of scale. In 2015, expect the majority of the world’s largest IT organizations to still be leveraging z/OS with its security, availability, and so on, supplemented with z/VM and Linux virtual servers as the hub of their computing utility.

Though mainframe customers have several areas of legitimate concern, none of the mainframe’s pain points are life-threatening. The top three complaints are high cost, a shrinking skills base, and lack of applications. Customers indeed pay a premium price to use the industry’s premier server. IBM’s pricing model is broken, especially for software. Sub-capacity pricing and specialty application processors (e.g., zAAP) are certainly two steps in the right direction. Yes, some ISVs charge outrageous license fees and many small-to-medium-size businesses will find mainframe entry costs prohibitively high, but the utility model will lessen these pain points, too. Emerging utilities will discover that when people costs are accurately assessed, the mainframe offers the lowest cost per user, similar to how jumbo jets become economical when passenger loads are high.

It’s no secret that people with mainframe skills are a shrinking population. For example, one shop with 60 people in their mainframe group will lose half to retirement during the next five years. IBM is also responding to long-term, skills-building concerns by signing up more than 200 professors and 150 universities worldwide to equip students with mainframe skills, and simplifying many of its operations tasks through its Health Checker and use of newly acquired Candle technology. OK, but for the last four decades the IT vendors and customers have relied on their own training programs to produce mainframe skills, and this will continue.

Since IBM’s “WebSphere everywhere” strategy encompasses z/OS, any apps written to the J2EE standard can also run on the mainframe. This has the potential to alleviate concerns that application development takes too long on z/OS, and allows application deployment on the platform that best matches quality of service needs. Longer term, look for IBM to make z/OS available on every developer’s desktop, which will increase mainframe awareness among application architects.

For the past 10 years, the mainframe has suffered from a poor image among the misinformed, and it’s an unknown among the uninformed. z/OS and the new System z9-109 have the right ingredients for a computing utility, and partitioning is the secret sauce. I predict that IBM will eventually increase awareness beyond the mainframe community of all these advantages, and by 2015 the mainframe will be celebrating 51 years of unparalled accomplishment.

Grace and peace.