isham research
IBM's 7 October mainframe announcement (GA2 of the z990) was widely expected, having been signalled in the original z990 announcement. A pricing announcement was also expected - but nothing warned of the biggest surprise. The mainframe is back!
Absolutely true. The first line of text on the first foil in each of the (separate) presentations given to industry analysts and business partners at the end of August 2003 says:
Several foils repeat this supposedly obsolete term. The mainframe is back.
Does this make sense? In a world that increasingly regards information technology as a utility function, it's perhaps worth considering what that means. No major organisation, for instance, has a dining room for each department - there's a communal cantine facility for each building or campus. Similarly, you don't find air conditioners in each department, or little Honda 4-stroke power generators under each PA's desk. The same goes for elevators, company cars, etc. etc., so why should IT be different? Why, indeed, do so many user departments have their own servers?
In many ways, the explosion of distributed computing (and especially using the Windows platform) is one of the greatest confidence tricks of modern times. Is it really cheaper to have every secretary in the company change a toner cartridge twice a month, even without the training and health & safety costs, than to have professional operators do it centrally with proper dust extraction equipment? Is it easier to build synergistic applications - for competitive advantage - on one system or thirty? Is it safer to have one disaster recovery plan, or forty? When true total cost of operation is examined - in a complete picture including peer support issues, training costs, disaster exposure, etc., a glasshouse (perhaps that term will also return) is always better for the bottom line - but that's usually the only place it's visible.
Viewing IT as a utility function doesn't mean it has to be outsourced to a utility company - and even if it is, it doesn't have to be off-site. But it does have to be flexible and competitive. IBM's product strategies over the past few years (and especially its espousal of Linux on the mainframe (sic) have vastly improved flexibility - but pricing has needed attention for some time. With this in mind, and quite apart from the readoption of the term "mainframe", IBM's August 2003 announcements must be very welcome to adherents of the mainframe:
A lot of the pricing initiatives are aimed at making IBM's mainframes more competitive against smaller systems without IBM having to give deep discounts. If a small workload is to go on a large machine, many "total system size" pricing algorithms make it prohibitively expensive. The traditional approach is to give a discount under Special Bid - but this means IBM loses out when the application grows. It's far better to license at the pro rata for a part of the machine, so charges can be increased as the workload grows. Given that mainframes have always hosted multiple workloads, it's amazing both that the original 1987 mistake was ever made, and that it has taken so long to try and correct it.