Revise and consent – but take your time about it
Since 1956, IBM – and therefore the world’s trade in high performance commercial computer systems – has been shaped more by a contract between itself and the US Department of Justice than by any other single identifiable cause. That contract is called the Consent Decree, a compromise reached to settle an antitrust lawsuit brought by the US in 1952. The lawsuit stemmed from IBM’s conduct in the accounting system business of the time, a business whose dominant hardware included card punches, tabulating machines and other unit record machinery. Justice viewed IBM’s market position as a monopoly that was built, maintained and perpetuated by means that violated the law. IBM disagreed, vehemently. In an attempt to force IBM to change its ways, Justice took IBM to court. Four years later, the battle was still going on.
Compete with itself
By that time, both sides realised a settlement – a compromise of some sort – might be the best solution for all the interested parties, which included not only IBM, its employees and shareholders, but the entire business community and US general public, for which Justice happens to work and represent. IBM equipment was so good, so prevalent and so likely to remain the standard in the US (and the world at large, although that was not a concern of the US government) that there was only one way to get a little competition into the business. IBM had to be made to compete with itself. Coming up with a plan to do this was ingenious but also very simple. Until 1956, IBM didn’t sell its unit record equipment. Customers could only rent the machinery. This allowed IBM to keep its hardware on rent for very long periods. All IBM had to do was to maintain it flawlessly, which it did. And from time to time, IBM improved a machine or introduced a new piece of tab card apparatus. IBM was so far ahead and its penetration of the market so great that no company could take advantage of any possible lethargy on IBM’s part to bring out a machine that was more attractive than IBM’s offerings. What IBM agreed to do in 1956 was to sell as well as rent its machines, to sell spare parts so customers (or their agents) could maintain the machines and never to sell or rent any equipment it once sold to a customer and later repurchased. In any event, by accepting the compromise, IBM ended the certainty for continued litigation by the government and avoided the possibility of an even more dangerous victory by the Department of Justice. Now IBM wants all these restrictions removed. Companies whose business it is to buy, sell, lease or maintain IBM machines are understandably opposed. But the main interested party, the users, don’t have an advocate. The US Department of Justice has said that it will not oppose IBM’s efforts. In the government’s view, there is presumably no threat to the American business community lurking in IBM’s desire to abrogate the restrictive contract it signed 38 years ago. We hope the process takes a while. We’d like to feel that IBM’s potential opponents, including the Justice Department, have given the matter the very serious consideration it deserves.From Infoperspectives International, July 1994, published by Technology News Ltd, 110 Gloucester Avenue, London NW1 1JA. Copyright (C) 1994 Technology News Ltd.
Taking stock of IBM’s first foray into RAID storage for its mainframe users
At long last! IBM’s RAID disk arrays, using the venerable RAMAC moniker (borrowed from IBM’s first commercial disk drive), have been announced. They are somewhat different from the products IBM had once suggested it would offer, but not radically so. The disk subsystems come in two styles. One uses regular old 3990-3 or 3990-6 cache disk controllers but replaces 3390s with RAID modules. The other includes a cache controller attaching directly to a channel. Both use disk arrays that are small and easily replaced or upgraded – four-drive units that yield 5.67Gb net storage. IBM calls the arrays drawers. If the disks do not hit their initial sales targets, IBM will cha
nge its drawers and try again. The changing of the drawers to double their initial capacity, which IBM and its customers may well need, is to occur in the next several months. No date has been given; the upgrade will be announced just as soon as IBM figures out how to put more data in and not clobber performance. Each string of disks, whether for the 3990-attached or channel-attached version, can include from two to 16 drawers, for a total effective capacity of 90.7Gb. A 3990 supports two strings, while the three model 9394 controllers support one string. Two of the channel-attached controllers are dual units, differing in their power requirements, channel attachments and maximum cache, the third is a quad controller with even more channel attachments and cache.
Rushed its initial announcements
The 9494-001 with 64Mb cache and four channels (Escon or parallel) costs $135,000 plus $415 a month for maintenance. A 64Mb 9394-002 costs $150,000 plus $425 a month; it can link to four Escon or eight parallel channels (or for $10,000 more, four Escon plus four parallel). An upgrade for either the model 001 or 002 to 256Mb cache is $81,000 plus $150 a month. On the 002, there is also a 1Gb cache option at $322,000 plus $690 monthly maintenance. The 9394-003 quad controller costs $215,000 plus $550 a month. Its base configuration includes 128Mb cache and eight channels (Escon or parallel). A 512Mb cache option is $162,000 plus $300 a month; a 2Gb cache upgrade is $644,000 plus $1,380 monthly. A channel option that provides eight Escon plus four parallel attachments costs $10,000. The disks for all the channel-attached controllers are designated 9395-B13 and cost $30,000 each plus $50 per month maintenance. The configuration of the 3990-attached machines is much simpler. A base machine includes a model 9391-A10 – a $50,000 (plus $55 maintenance) rack – and two model 9392-B13 drawers at $30,000 each (plus $50 maintenance). The two B13 disk arrays are physically identical but apparently use different microcode. IBM did not announce a plan to change the microcode of disks it has sold. Until IBM offers a conversion option – which we expect it will do, eventually – users who buy one type of array will be unable to move their disk drawers to the other type. IBM has made it clear that there will be B13 drawers with twice the current net capacity in the near future. Announcement early next year with deliveries by July is seen, but if IBM can get there sooner, it will. IBM has also said that IBM Credit Corp will offer leases that take disk upgrades into account. This strategy suggests that IBM rushed its initial mainframe RAID announcement as a defensive measure. IBM has not addressed the big issue: how will early users move the critical data they put on their first IBM RAIDs to dual-capacity RAID arrays? The answer for many prudent customers will be They won’t, because they will wait for the potentially addictive product to stabilise before making a strategic commitment. IBM’s competitors do not appear to be in trouble yet and might not be if IBM has teething problems. Insiders at StorageTek sound pretty confident in their Iceberg and assert that it will outrun IBM’s arrays in demanding applications. Nevertheless, early Iceberg users will face the same problem as those with the first IBM RAID disks. StorageTek will move from the old Hewlett-Packard Co ESDI disk drives to much more cost-effective and compact SCSI disks as soon as it can. This will make initial disks less desirable and probably create migration annoyances for customers who want to move up. Users buying the disks will also have to ponder whether early Icebergs will have acceptable residual values. EMC Corp has no RAID, but it does have a large and growing customer base for its subsystems. For now, EMC is offering some users double capacity systems (with disk mirroring) on a free six-month trial, at the end of which users can buy the extra capacity or, if they feel the equipment is sufficiently reliable, return the unwanted drives. We believe EMC will have to unveil a RAI
D product – or a way to replace the regular drives in a Symmetrix with RAID modules – by early next year for delivery during the second or third quarter in order to sustain its high growth. – Hesh Wiener
