Mainframe Performance Monitoring

The early '90s were also a challenging time for IBM, which had lost its way since taking over the mainframe world in the '70s. In 1992, IBM Senior VP Richard Gerstner was tapped to turn things around, but his personal energies went into a decline which turned out to be a new ailment, Lyme Disease. Richard's brother Lou, then head of RJR Nabisco, picked up the challenge instead and is credited with bringing IBM roaring back to life, focusing on its core competency of huge mainframes running ancient COBOL applications in business data processing. Thus did a nearly invisible tick change computing history.

The problem with tools that remain the same for decades is that the problems they're addressing can change in a way that demands upgrades in the tools but doesn't explain what they should be. Another contributor to IBM's recovery was a small software house named Programart, founded in 1969 by three men, including my old Apollo colleague Bob Morse -- though I didn't know that until I'd joined the company.

Programart's product, Strobe, investigated the performance issues of mainframes by counting how often certain software modules were run, especially those that had to await completion of disk drive operations. As corporate data processing experienced explosive growth, inefficient modules and disk usage strategies became huge time wasters, often negating the benefits of upgrading to new hardware. Strobe reports pinpointed where software upgrade efforts had to be concentrated -- and that's where Easel on both PC platforms (OS/2 and Windows) came in, as a medium for designing each customer's measurements and presenting the performance monitoring reports in graphical form.

By 1995, Easel Corporation had lost most of its tech support energy, and ambitious customers like Programart needed to bring in Easel hired guns like me. After a few years of upgrading the Easel front ends which obtained their performance data by capturing what mainframe Strobe code was preparing for dumb 3270 terminals (an art colorfully known as "screen-scraping"), we decided to adopt the Internet approach, receiving XML files over the intranet and using XSLT to convert them to HTML pages for display.

The widespread anxiety over what Year 2000 was going to do to software, especially old COBOL programs that had allocated two-digit fields for the year portion of 20th-century dates, caused a boom for such companies as Compuware, which used their mountain of cash to buy out Programart's founders and make Strobe part of their consulting arsenal. When the "Y2K" problem turned out to be a wet squib, Compuware's interest in Strobe and the Programart people declined gradually. After a strange couple of years maintaining COBOL code for the mainframe part of Strobe by creating program patches in hexadecimal and issuing them to our customers as ZAPs, I was put out to pasture.