Milking the Mainframe: Exorbitant Software Licensing Costs

Here’s an IT history lesson we seem doomed to repeat: Vendors killing their own business to protect their margins.

The history: Once there were Apples, made by Apple, and there were Windows PCs, made by anybody who wanted to license the operating system from Microsoft. Apples carried a premium price. Apple didn’t let anybody else make systems to run its OS. After all, why should the mighty Macintosh maker let a bunch of cheap clones come in and start a price war?

You’ll also recall what happened as a result. As soon as Windows became a remotely viable technology option, sales of ever-cheaper PCs took off. Microsoft got rich; Apple got niche.

The lesson: Margins are not the name of the song in IT. If the farmer wants 100 gallons of milk per day, he’d better not try to get it all from one cow. It irritates the cow.

The repeat: The victim this time is not Apple but the mainframe, which has fallen further and further from favor. Users’ biggest beef? Exorbitant software licensing costs. The worst of it has been that when you buy a bigger mainframe engine, you pay more for the same software, regardless of the number of applications or users or copies or any other rational measure of software usage. Imagine applying this licensing policy in the PC world: Microsoft Office would cost $50 if you run it on a Pentium II, but $150 dollars if you buy a Pentium III.

There may be all sorts of justifications for the high cost of mainframe software. However, vendors should stop justifying and look at the bottom line. Mainframes aren’t extinct (see “IT-Rex?” Page 224), but they’re endangered (at press time, Amdahl announced it’s dropping out of the business). And mainframe purveyors should be kicking themselves over that, because it didn’t have to happen.

The biggest communal leap off the mainframe was the client/server revolution, and early client/server was an architecture with all sorts of manageability, scalability and reliability problems. You have to figure that if the price of mainframes had been bearable, the corporate world’s exploration of client/server would have been much, much slower. And ultimately a lot more companies would still be running MVS today.

In September, when IBM announced its latest generation of mainframes, the wall finally came down a bit. IBM joined hands with BMC Software, Computer Associates, Compuware and hosts of other mainframe software vendors to declare “more flexible licensing policies,” meaning that users will get charged for what they use instead of how big their mainframes are.

Good, but a bit late. And because it’s late, the announcement is probably not radical enough. Nobody is going to confuse flexible with cheap.

The so-called new economy is nothing if not transaction-heavy, and in many respects mainframes are still the most attractive platform for that sort of processing work. If the software were to become dramatically cheaper, I think we would see a bonafide mainframe revival. If not, well, I think the remaining cash cows are mighty tired of being milked.