IT Management


To the unwashed, the cryptic title to this column is intentional: I deliberately used the popular abbreviations for application development and maintenance—APP DEV/MAINT—not to shorten the text to a Twitter-able size, but to immediately get to a point. We’re lacking comparative data to help illustrate something we intuitively already know, but haven’t bothered to quantify and verify up until now: Application development and maintenance are far less costly in a mainframe environment than in a distributed environment.

This point was underscored for me in an email from a reader. The writer, whose handle of “solutions architect” at a medium-size Midwest insurance company asked me for some statistical data.

Noted the fellow, “For all the hype about IBM lowering costs for TCO and TCA for the mainframe, I noticed a very important cost component that isn’t being covered: the cost to develop and maintain applications for a distributed environment vs. the mainframe.”

Had I not met this gentleman before, I might have thought I was being set up for another one of those “distributed computing trumps mainframes because …” observations. But he went on to observe, “The distributed environment is much more complex than the mainframe environment. Developing on distributed platforms requires staff with extensive expertise and a very broad knowledge base. Has anyone even looked at the [differences in application] development costs between distributed vs. mainframe? I can get something to production on a mainframe in a more stable way and faster with mainframe developers than I can in a distributed environment, and I can generally use developers who have fewer years of experience and knowledge than what would be required [of developers tasked to do the same work in a distributed environment].” The fellow went on to say that application “break/fix” also was easier in the mainframe world: “Mainframe application fixes are faster and quicker because there’s far less complexity to work through. Granted, hardware and software costs are part of the total, but the real costs are the development and maintenance costs … and there’s a world of difference between distributed and mainframe.”

I sensed a tinge of frustration in this note, so I followed up with a phone chat. He told me the CIO at his company had decided to pursue a five- to seven-year plan back in 2000 to migrate off the mainframe. The project, he reported, had now been extended to a 10-year timeframe, mostly for reasons of cost and resource availability.

Why the plan to leave the mainframe? He suggested that the proliferation of insurance policy management and maintenance applications for distributed computing environments, aimed at small- to medium-size providers, was communicating the impression to management that mainframes were no longer needed. He disagreed with this perspective, of course, noting that the I/O-intensive processing required nightly to match policy exposure to cash reserves on a continuous basis would be very difficult to provide in a distributed environment.

He said he had some older data suggesting that any company that was using 400 MIPS or more of mainframe compute power to handle workload would probably never be able to migrate that workload off the mainframe and onto distributed environments—at least, not without a lot of pain. But, he lamented, that study was now a few years old. He said he needed new data points that would underscore the lesser cost of application development and maintenance in the mainframe world—something intuitively obvious only to mainframers.

His is a valid request, and it’s even more timely in light of the struggle confronting most companies today to discern the right staff size for distributed computing environments. Over the past month, I’ve seen a half dozen articles in the trade press in which analysts and pundits were wringing their hands over the meaning of “lean IT staffing.” Estimates for IT staff requirements are all over the board.

In the mainframe world, things are different. We have well-defined staff descriptions that reflect a time-honored order of battle. As a rule, seasoned mainframe managers know, based on their application set, how many bodies they really need. This isn’t true in the distributed world.

To try to pin down the reality here, shortly, we will be launching a series of polls on the C-4 Project Website ( with a goal of capturing sufficient data to enable a back-of-envelope comparison on the differences between APP DEV/MAINT costs in different environments. Please join in, and we’ll report the results here.