The AS/400 has been largely ignored by IBM of late. Nick Enticknap finds that the old soldier will never die but may simply fade away
IBM has four distinct server ranges, which seems at least one too many. The Wintel (xSeries) and Unix (pSeries) families seem safe and it is unthinkable that IBM would ditch its mainframes (zSeries). They still generate at least a quarter of the company's revenue - so nobody is worrying about that axe yet. This leaves the AS/400, or the iSeries as we must now learn to call it, looking very vulnerable.
IBM Computer Users Association (CUA) chairman Ray Titcombe, himself an AS/400 user, is one who fears that IBM's rebranding announced earlier this month could be seen as the road to oblivion. "It is important that the message gets through that they are not trashing the AS/400," he said.
Software company Geac, a long-term supporter of the AS/400, is quite confident that IBM has no axe behind its back. "In five years time we will be having the same conversation," says Bernie McHugh, Geac's head of channel operations. "The AS/400 will be here for a considerable number of years yet."
IBM, naturally, is singing the same song. "The AS/400 will certainly survive, because it makes so much money," declared Paul Fryer, IBM Enterprise Systems Group marketing communications manager. "Counting hardware, software and services, the AS/400 business is the same size as Sun." That makes it a $15bn (£9bn) business.
The AS/400 has other things going for it, too. It may not have the ISV support that Windows and Unix have, but there are over 40,000 applications out there already, more than 10,000 of them for Lotus Domino.
And it has a very large customer base. IBM says it has sold more than 650,000 AS/400s over the product's 12-year lifetime. Gartner estimates that over 500,000 AS/400s are currently in use, showing that the customer base has proved exceptionally loyal.
The AS/400 shows up well in surveys. IDC, for example, has found that the AS/400 had a significantly lower total cost of ownership than comparably configured Hewlett-Packard and Intel/ NT servers.
In addition, the AS/400's unique architecture has shown itself to be every bit as future-proof as IBM has always claimed. The AS/400 was born into a world of green screens, mainframes and master-slave communications. It has survived the client/server revolution and is now happily positioned as an e-commerce platform. All its major competitors at the time of its launch -DEC Vax, Hewlett-Packard 3000, Data General Eclipse, and so on - have long since vanished.
Despite this, the architecture is also a disadvantage as it is quite unique, totally unlike that of the other server brands, particularly in its treatment of all its storage devices as a single homogeneous pool (single level storage). This feature led IBM to bring disc storage inside the covers of the AS/400 five years ago, as this provided significant increases in performance both in running OS/400 applications and also those on the integrated PC servers under OS/2, Novell and Windows NT.
Unfortunately, this was when the rest of the world was embarking on the storage revolution that broke the tight server-storage bond and led to the emergence of storage area networks (Sans) and network-attached storage (NAS).
As a result, the AS/400 can only participate in Sans through slow adapters developed for the first-generation SCSI world that have not been enhanced since. IBM has promised to support Sans in the future, but further information on this was disappointingly absent at the iSeries launch.
Also conspicuously absent from that announcement was any further information about support for Linux on iSeries. All IBM has said is that Linux will be supported and that the method of doing so will be logical partitioning. As things stand, iSeries is the only one of the eServer ranges that cannot run Linux.
This absence of participation in two of the major industry trends, Sans and Linux, could be seen as ominous, suggesting that already iSeries is being put on the back burner.
Titcombe doesn't think so. He believes the iSeries management in Rochester "didn't want to waste good marketing material when most of the corporation was swamping the media with announcements." He added, "I believe that next spring Rochester will fight back. I think the San approach will come then."
But the real key to the future of the AS/400 lies in the eServer rebranding announcement. As this stands, it is just a marketing announcement. But it came in the middle of a chain of development that has already been going on for five years - and will probably take another five to complete.
The chain is the unification of IBM's server ranges. It started when IBM converted to Cmos logic circuitry for its mainframes in 1994. Since then, the base technology used in all the IBM server ranges has been the same and IBM has been steadily moving towards a position where the actual components are also the same.
Today, the processor chips used in the iSeries, pSeries and zSeries are made in the same factories and are 90% the same product. The 10% difference covers the layers used for the instruction set. The disc and tape technology is also the same and, more recently, this has been extended to sub-system level - products such as the Shark disc sub-system and the Magstar tape drive are available for all three ranges.
IBM has also started to move towards software convergence. The DB2 database is available on all four platforms, as is the Websphere Internet software.
All of this means that IBM already has a lot of R&D shared across all four server ranges, with obvious economy-of-scale benefits. The eServer branding is a recognition of what has already been achieved, rather than the end of the process. It is, as IBM enterprise server vice-president Stephen Murdoch said, "the next step, as opposed to the final point we're trying to get to".
So what is that final point? According to independent IBM consultant Barry Graham, "In five or six years, the same chips will be going in all the platforms so you can be running anything you like on all the platforms."
In other words, at that stage today's AS/400 customer looking for a new model will simply purchase an eServer - the iSeries branding will have completely disappeared as far as the hardware is concerned. This future eServer will run OS/400, AIX and OS/390 applications - all natively and concurrently, if required.
It sounds far-fetched but actually we are some way down this road already. In January, IBM announced Pase (Portable Applications Solutions Environment) for the AS/400. This is an integral part of OS/400 which provides AIX runtime functionality, and so allows AS/400 users to run recompiled AIX applications natively under OS/400.
Pase takes advantage of the fact that the processor chips in the AS/400 and RS/6000 are already exactly the same. They run in either AS/400 or RS/6000 mode, so, as IBM's Fryer explained, "The AS/400 chip can become an RS/6000 chip by flicking a switch." That switch is soft-coded so it can be flicked by a program and a Pase application does just that. The next OS/400 program to require that processor just flicks the switch back again.
The future that Graham foresees takes this principle a step further. Processor chips for pSeries, iSeries and zSeries boxes will be exactly the same and the applications software will tell the processor which mode it needs at runtime.
In this context, the debate about whether the AS/400 has a future becomes irrelevant. As an independent hardware entity it will disappear, but AS/400 sites will still be able to run their existing applications indefinitely. As Barry Graham puts it, "It will die, but this will not worry the users."