Ten years ago the mainframe was a dinosaur. The concept of a big centralised computer, run by specialist staff was being replaced by lots of distributed smaller computers. Developments first of minicomputers, then networking and finally PCs meant that the mainframe was becoming obsolete.
IBM was the major beneficiary of the mainframe era, but not the only one; Unisys, Honeywell, Siemens, Bull, ICL, etc. all did very well from the old centralised architecture. However the strength of IBM at that time created a significant variant, lead by Amdahl, of hardware manufacturers making systems compatible with IBM machines, to such a degree that they ran the actual IBM software such as VM, VSE and MVS. While this provided some competition for IBM, it also kept them on their toes. It also created an opportunity for the Japanese, Hitachi and Fujitsu, to enter the international IT market in a big way. This concentration on IBM software was so dominant that it more than anything caused the gradual demise of the non-IBM compatible mainframe. This was not due to any superiority of MVS, etc., in fact quite the reverse, it was due to shear market presence. The same can be said of Microsoft today in the PC sector.
By the beginning of the 90’s Unix was also making an entrance into the commercial world. Hewlett-Packard and Sun Microsystems made inroads into the mainframe market, and the Unix technology grew to such an extent that the "other" mainframe manufacturers redirected themselves towards it, but with limited success, squeezed between the size of IBM and the innovation of the specialist Unix companies.
Thus the mid 90’s saw IBM struggling alone in the mainframe market. The very term "mainframe" had become synonymous with the System/390 architecture and its derivatives. Even IBM themselves had become serious players in the Unix market as well as PCs. But then came the Internet. The Internet defines a future which is based on distributed access, not distributed computing. This in turn means that gradually more and more processing will return to centralised systems, and that means bigger, more powerful and more robust systems. That means mainframes! Of course H-P, Sun, etc. will go on improving scalability and partitioning of Unix systems, and others, including mainframe fallouts such as Unisys and the Japanese, will foolishly continue efforts to do the same with Windows NT clusters, but IBM are miles ahead at the moment. No wonder IBM love the Internet and are so keen to sponsor the OSS movement.
Everything would have been perfect for IBM except for one thing, the communications industry has been unable to deliver on the promise of cheap, high-speed connectivity for everyone. While the mobile phone phenomenon may well eventually be the major player in the Internet, it is today only a fringe activity. The need to get a foot in the door at the beginning has inspired all the telecommunication network suppliers to invest in mobile networks, at the expense of wired systems. Even cable, with its integration with TV, has not made any significant breakthrough in providing high speed Internet connectivity.
The direct impact of inferior Internet connectivity has been a much slower take up of e- commerce than was anticipated. There are of course other factors such as the imbecilic waste of investment on dot.com disasters and the amateurish efforts to integrate the new systems with the established "legacy" systems, but arguments about the relative merits of .NET versus Java are not as critical as communication problems. Thus IBM have only be able to use their mainframe strengths to partially dominate the market, but not to create a stranglehold. This then is buying time for the others to play catch-up.
It is fair to assume that despite the relatively slow take-up, the next decade will see big investments in e-commerce continue. Thus all the current players will continue to strive to gain predominance, which means that they will be developing Unix, Linux and Windows machines which are addressing the key mainframe strengths of today. In the role of large scale systems this means an increasing emphasis on scalability, reliability, ease of management, power consumption, cost of ownership and partitioning. At the moment the mainframe, and that means IBM, wins on all counts. Can they keep this advantage and if not how long can they and what will they do to compensate? It is a fascinating scenario.