Research tools - laptops for the Internet and handhelds for mobile usability testing

isham research

Comment: Carving up mainframes - has IBM missed the boat?

Originally written on 23 February 1999 - updated in January 2008 for XHTML
conformance and some keyword changes. Text unaltered.

With Hitachi Data System's Trinium (sorry - they've now taken the web pages down) and Comparex's M3000 announcement, all of IBM's plug-compatible vendors have now defined ways of carving their large systems up into multiple systems with separate serial numbers. IBM has announced no equivalent capability and no current plans to provide one - so for at least the next two years this feature will only be available to Amdahl (shipping since 1998) and Hitachi (shipping in 3Q99) users.

Is this important? Yes. Probably very important. The problem is that IBM has lost control of software pricing on System/390 mainframes - only a very few of the independent software vendors (ISVs) have followed IBM's pricing initiatives over the last few years, and most are still stuck on some version of simple sized-related 'graduated charging'.

As footprints get larger, this is increasingly inappropriate. Hitachi Data Systems' roadmap for Skyline Trinium promises over 4000 MIPS in a single footprint, and so it made a great deal of sense for HDS to follow Amdahl's lead and make Trinium divisible into separate servers.

IBM refuses to recognise such divisions, insisting on charging for its products according to the total capacity of the system. Since all that many PCM users buy from IBM is the operating system and its support packages (DFSMS, DFSORT, etc.) this doesn't matter and may even be beneficial - otherwise users would have to jump through hoops to establish PSLC aggregation qualifications.

The ISVs reacted very differently. Both Amdahl's Multiple Server Feature (MSF) and HDS' Virtual Server Facility (VSF) give them exactly what they need; a way to know that users are not deriving more benefit from using their products than those users are paying for - simply, cheaply, and without code changes. As time has passed since Amdahl's original announcement, more and more ISVs have quietly indicated their willingness to acknowledge MSF - with HDS' announcement, Computer Associates have signed up in public.

Since ISV software now accounts for more than half of most users' software bills, IBM's software pricing practices are becoming less influential than ever.

IBM is now at a disadvantage. Not only does its hardware not offer a comparable feature, but its middleware (e.g., DB2) must struggle against ISV products whose vendors do not insist on licenses being paid for capacity unavailable to the application. IBM seems to be trying to respond via a new 'Application Level Charging' methodology - a system nothing like as simple as simply partitioning off engines, and one unlikely (again) to be adopted by the ISV community.

There's another aspect - upgrades. Until now, those who would upgrade a mainframe have faced two easy tasks and one monumentally hard one - choosing the hardware, pricing the IBM software, and then negotiating with their ISVs. With MSF or VSF, this can be a thing of the past - where an extra engine is added to one of the independent servers, there is no need to negotiate with those ISVs whose products are used in other, unchanged, servers.

isham research Home Page and COntact Information