IBM Corp has and almost always has had absolute control over the AS/400 disk market. The fact that no third party has ever successfully taken on IBM and beaten it in the AS/400 market is a testament to the cleverness of IBM’s disk designers in San Jose. They got RAID-5 to work first in the mid-range, years ahead of Hewlett-Packard Co, Digital Equipment Corp and others. They make some of the best and fastest disk drives in the world. But IBM San Jose also makes AS/400 disk subsystems that usually run a little weak on microprocessor power, low on cache memory and short of other bells and whistles. So IBM’s marketers deserve some credit for the company’s success, too. Year in and year out, they have won most of the disk deals in the AS/400 world, sometimes with less than exotic products. IBM’s strength in the AS/400 disk business and the almost complete lack of interest by third parties (at least in recent years) may indicate there’s more that’s different about the AS/400 market than at first meets the eye, something a bit ominous to those of us who are fans of the AS/400.

Not bode well

The fact that no third party has a significant share of the AS/400 disk market – non-IBM disk sales account for about $125m of the $1,800m of disks sold in 1995 – may show just what disk integrators think about the potential for growth in the AS/400 market. It could just be that disk makers, who could get into the AS/400 market just as easily as the mainframe or open systems markets, don’t think there are good prospects for sales and profits there. If they are right, that does not bode well for the AS/400. Both IBM and those who have tried to take on IBM in the AS/400 disk market (and failed) might say that it is IBM’s strong position in the AS/400 world that makes it nearly impossible for them to develop and market products for the AS/400. This, of course, is no excuse. IBM had much more control over the mainframe disk market, and now EMC Corp sets the pace there. EMC took the long view and instead of copying IBM’s mainframe disks, it developed inexpensive alternatives that provided the same function and more. The company’s flagship Symmetrix disks (which are now supported on AS/400s) not only look like IBM disks to the software, but they are faster than IBM’s disk subsystems, largely because of great gobs of cache memory. Symmetrixes boost system throughput and often allow customers to forego a very expensive mainframe CPU upgrade. IBM’s competitors have tried the same tactic in the AS/400 market, and at times they have been successful in gaining a foothold in the AS/400 market. During the late 9335/9332 and early 9336 generations, disk subsystems from EMC, IPL Systems Inc, XL/Datacomp Inc and a few others usually used faster disks, better controllers and lots of cache memory. They were indisputably faster than the IBM alternatives. The third parties got some of the action at AS/400 sites, but by and large, IBM still sold between 80% and 90% of the disk capacity the AS/400 market bought each year. IBM has pretty much killed the third parties with its second and third generation 9337s, which have RAID-5, enough cache memory to get by and reasonably peppy controllers. The competitive situation is made even more difficult by the Advanced Series, which now supports completely internal disk subsystems for 9406s as well as deskside processors.

By Timothy Prickett

None of the mid-range vendors is selling internal disk subsystems for the AS/400, which is what AS/400 customers are almost exclusively buying today. This switch from external to internal disks took place very quickly, and against what is better judgement. (We believe that competition is good for AS/400 customers). In 1994, the majority of disks bought by 9406 customers were external 9337s or clones of them made by third parties. In 1995, about 70% of the disk capacity sold to AS/400 customers was internal disks; IBM got most of the remaining external disk sales, too. Despite the tough competitive environment in the AS/400 arena, we still believe

that none of the mid-range disk vendors would walk away from the AS/400 disk market if it weren’t for two other mitigating factors: they can make a lot of money in the Unix disk market, and they can make even more selling disk subsystems for personal computers servers. According to estimates made by disk analysts at International Data Corp, in Framingham, Massachusetts, the AS/400 disk market has remained essentially flat since 1993. IDC expects that it will remain flat between 1996 and 1999, too, at about $1,700m a year. To keep disk revenue flat like that, IBM will have to sell about 20% more capacity a year (some will be additional capacity to existing AS/400 users, some will be disks sold in conjunction with whole new systems) while prices drop about 20% a year. Compare this to the Unix disk market, where disk subsystem prices are anywhere from 10% to 20% lower than for equivalent AS/400 technology. IDC says the worldwide Unix multi-user system disk market has grown from $2,900m in 1993 to $4,800m in 1995. Analysts at IDC expect that the Unix market will grow to $7,400m in 1999, which will make it six times larger than the AS/400 disk market. Similarly, IDC expects that the personal computer server market, which has grown from $1,600m in 1993 to $3,200m in 1995, will double again to $6,500m by 1999. It doesn’t take a disk head to figure out where the action is in the mid-range. (Just to make your head spin: the single user personal computer and workstation disk market is about $22,000m a year, and will double by 1999, too. There’s a reason why IBM wants to be a big player in the OEM disk business. It may be easier for IBM to be an OEM disk maker than to fight EMC head-to-head in the Unix and PC server markets).

Divide-and-conquer

The other major trend in the mid-range disk business doesn’t have to do with money, but how disk capacity is organized at end user companies. Subsystems from all the major players now support multiple concurrent system attachments as well as a wide variety of file system types. The Symmetrix 3000, for instance, can support up to 32 individual AS/400s. Or, if customers want to consolidate the data storage for their AS/400s, personal computer and Unix servers, they can put all their files under one big Symmetrix, too. (Storing different files types on a single disk subsystem does not mean that all systems can use all files. They can’t. Having one big storage subsystem just lowers the cost per unit of capacity). IBM has only recently offered disk subsystems that support multiple environments with its RS/6000 7137 units, and even then the subsystems only support AIX, HP-UX and Solaris. If IBM comes to the conclusion that its divide-and-conquer strategy in the mid-range disk business is costing it sales to EMC and others, it may announce large disk subsystems suitable for big AS/400, RS/6000 and NT customers. But we haven’t heard of any such development projects at IBM San Jose: we don’t expect any, either. From The Four Hundred, February published by Technology News Ltd, Copyright (C) 1996 Technology News