Microsoft and Micro Focus Go After Mainframe Apps

by Timothy Prickett Morgan

Last week was the fortieth anniversary of the launch of the venerable IBM System/360 mainframe, the mother of all mainframes and in some ways the mother of all modern computers. It was no coincidence that Microsoft and its COBOL partner Micro Focus, which provides COBOL compilers for mainframes, Unix boxes, Windows gear, and Linux servers, were touting their strategies to move mainframe applications off mainframes to try to steal some thunder from IBM.

Micro Focus was founded in 1976 and really came into vogue with the first wave of application migrations off the mainframe, when Unix servers became a viable alternative to big IBM and compatible mainframe iron, in the late 1980s and early 1990s. But it really made a name for itself in the mid-1990s, when it was offering both Unix and Windows COBOL environments that allowed applications from IBM mainframes and certain proprietary platforms (like the AS/400) written in that language to be moved to this much less expensive iron. Micro Focus has seen its ups and downs over the years. In 1998, Micro Focus acquired Intersolv, an ODBC developer, for over half of a billion dollars and formed Merant, which was to become a powerhouse in application development. In August 2001, after many trials and tribulations, Micro Focus was spun out of Merant to some venture capitalists for a piddling $62.5 million.

These days, Micro Focus sells a COBOL compiler environment called Express and a related transaction processing add-on called Enterprise Server. There are three different flavors of this COBOL environment: one that runs on mainframes (Mainframe Express), Unix boxes (Server Express), and Windows platforms (Net Express). Enterprise Server runs mainframe-style CICS programs as well as COBOL applications with J2EE extensions and Web-extended COBOL on Windows, Unix, and Linux platforms. It is basically a COBOL transaction processing rehosting environment with Java and .NET extensions.

So why are companies like Microsoft and Micro Focus still talking about chasing the mainframe's legacy applications two decades after the idea was first seriously considered? There's money in it, for one thing. Maybe not the $1 billion software opportunity that Micro Focus thought there was when it created Merant, just as the Y2K crisis was getting underway and IBM itself thought it would lose half of its mainframe installed base. (I saw the internal documents, and that was the projection.) IBM may have lost a lot of footprints, but the best estimates are that IBM still has around 20,000 mainframe footprints and around 10,000 mainframe customers. The top 2000 of these customers account for roughly 80 percent of the approximately $4 billion in mainframe sales IBM booked in 2003.

That may sound like a lot of money, but some historical perspective might shock you. The mainframe server revenue stream (not including peripherals and memory, but just the core boxes) was $13 billion at IBM alone in 1991, the year that the mainframe business started to get hit by Unix. (Amdahl, now part of Fujitsu, Hitachi, and Siemens all sold billions of dollars of IBM-compatible mainframes, too. And Unisys and a slew of other vendors sold their own mainframes with their own operating systems.) By 1995, mainframe hardware sales dropped to $6.5 billion, when Unix was absolutely credible as a mainframe alternative and Windows NT was becoming an alternative. Windows and Unix have since that time tag-teamed on the IBM mainframe base, forcing down the striking price of processing capacity from around $70,000 a MIPS with the ES/9000 9021 generation in 1991 to less that $1,800 per MIPS with the largest of the current zSeries machines, the z900 "Freeway" and the z990 "T-Rex" servers, today. So why chase the mainframe?

Because IBM has a monopoly on the mainframe software and it can and does charge very high monthly software licensing fees for it. The typical high-end mainframe shop spends anywhere from tens of thousands of dollars to millions of dollars per month. The zSeries hardware, while exorbitantly expensive compared to cheap X86 iron, is not the pain point today for mainframe shops.

That IBM software stack--z/OS, DB2, CICS, and a slew of related and necessary--is what have Microsoft and Micro Focus rubbing their hands with glee. But getting mainframe customers to switch, after 20 years of promises, has proven to be quite difficult, and there is no reason to believe that these two partners in a Mainframe Migration Lift-Off marketing program will have any better luck than any other program I have seen in dislodging mainframe shops from IBM hardware and software. The fact is, a small percentage of any customer base is always dissatisfied and ready for a change, and these marketing programs can only give them a little push. And the most dissatisfied IBM mainframe customers are not the large shops with thousands of MIPS, but those with smaller and usually older systems at midrange companies with tight budgets. To put this in perspective, Micro Focus says that it has performed 270 mainframe migrations. That is a good business, but it is not a mass migration.

The "lift and shift" campaign that Micro Focus has launched to sell its Enterprise Server and the CICS-like Mainframe Transaction Option component on Windows, Unix, and Linux platforms has substantial backing from Microsoft, which is an important change. But the hard work is still the same. Micro Focus and its partners have to use its tools to convert Ada, COBOL, mainframe assembler, PL/I, RPG, and other mainframe languages to Micro Focus COBOL, convert CICS transaction middleware to the MTO piece of Micro Focus Enterprise Server, and DB2, IMS, and other mainframe databases have to be converted to Oracle's Oracle9i, IBM's DB2 8.1, or Microsoft's SQL Server 2000. This is difficult work, and unless Microsoft actually pays Micro Focus to pay customers to make the move, the appeal is going to be limited.

So when Microsoft and Micro Focus claim, as they did last week, that they can redeploy mainframe apps to Windows running the Micro Focus software stack, and cut costs by up to 50 percent, that won't fall on deaf ears, but these mainframe shops have heard that before. Many times.