How to convince the world that we have the best servers for e-business? Put an 'e' in front of them! How to get rid of the old-fashioned, proprietary image of mainframes and AS/400? Give them completely new names!
The iServer (nee AS/400) announcement on the same day rather bears this out. The substance of this announcement was extremely small. No new models, just existing ones with an 'i' in front of them. No significant new functionality, even where it is desperately needed; the long-awaited announcements on Linux and San support were conspicuous by their absence.
Considered purely as a marketing announcement, it is a good one; IBM must have found itself a new agency. No buzzword-generated phrases like 'Modular Storage Server', which mean very little, and tell you nothing. Instead, a nice snappy unifying brand name; 'eServer', which puts across the message so clearly one wonders why nobody had thought of it before. (In fact, one company claims it has; whether it convinces a court of that remains to be seen).
But the rebranding is actually much more than a clever piece of marketing. It both recognises the convergence of IBM's server brands that has been developing over the past six years or so, and promises that this convergence will continue in future. This has profound implications for the future of user investments in AS/400, RS/6000, and System 390 applications and resources.
One example will illustrate both what has been happening, and what will happen in future. In the middle of 1994, the AS/400 and RS/6000 used very different processors, with different word lengths, as well as different hardware architectures. System 390 was not only different again from both of them, but also used different technology - ECL instead of Cmos.
Today, most of those differences have vanished. All three now, with the zSeries launch, use 64-bit addressing. All three now are built from Cmos logic circuitry. Indeed processors for all three ranges roll off the same production line. The zSeries processors are 10 per cent different from the other two, because of the instruction set. The processors used in the iSeries and pSeries are identical, and include the instruction sets required by both.
In mid-1994, the idea of any of the processors running an application developed for any of the others was a pipedream. Today, this actually happens; AS/400s can run Aix programs natively using the Pase (Portable Application Solutions Environment) capability.
It is not too difficult to project ahead and imagine processors being developed that can run all three environments at the same time. And once that happens, the physical distinction between an AS/400, an RS/6000 and a mainframe will have vanished. The iSeries, pSeries, and zSeries labels will then also vanish, at least from the front of the boxes. A user of any of the three wanting to upgrade will just go out and buy the latest eServer.
zSeries The most important feature of IBM's new zSeries is not the 64-bit addressing, valuable though that will undoubtedly be. It is the introduction of new workload based software pricing.
Traditional capacity based software pricing has been holding back user recognition of the merits of the mainframe as an e-business platform. IBM itself has recognised this, and come up with a few halfhearted expedients such as usage based pricing (MULC and later ULC), and new wave application pricing (NAGE), but neither began to address the problem. This was that mainframe software was far too expensive relative to the Unix and NT competition for small and pilot workloads; users simply could not justify the investment. IBM has now come up with a radical solution; to decouple software pricing from hardware capacity completely. On zSeries it will relate to the use the software makes of the machine's resources, varying dynamically.
The company has taken the logic of this move a step further by introducing capacity on demand pricing. Now hardware pricing will no longer relate to the capacity of the machine either.
Both steps do more to make the mainframe attractive for e-business than all the Java, WebSphere, and Linux announcements of the past three years put together. IBM has recognised two essential features of e-business:
workloads vary enormously in size; and,
people want to start small and grow large, and pay for the costs of that out of the revenues the e-business generates, rather than up front.
The virtual server facility ties in with this. It is a substantial extension to today's logical partitioning, allowing thousands of 'virtual servers' dynamically changing in size as the workload does.
These changes were essential if IBM is to make the mainframe's undoubted technical superiority count in the rapidly evolving e-world. We need to see the fine print, but the broad outline is very much the right shape.