Jump over left menu
Robb Wilmot: The ICL Story
Originally published in the November/December 1981 issue of Electronics & Power, the Journal of the Institution of Electrical Engineers, Volume 27, No.11, pp. 788-795. Republished by ICL as a publicity item
Britain's only world-scale computer manufacturer, recently just saved from bankruptcy, is now under an enterprising new management that is committed to re-establishing the financially-troubled company at the forefront of world computer development. What is the history of ICL, what lies behind its present financial troubles and what are its prospects for the future?
In the 13th year of its often apparently ill-starred existence, International Computers Ltd. (ICL), Britain's only world-scale computer company, has seemingly justified the forecasts of its many longstanding critics by at last reaching the very brink of financial ruin. To those who have always regarded ICL as the ill-conceived product of an unjustified and politically motivated amalgamation, and who have found much to criticise in its business record ever since its 1968 formation, this year's announcement of the company's virtual bankruptcy clearly came as no surprise. Even when the Government then baled out ICL with a £200 million guaranteed loan, and backed up its rescue plans by appointing a tough new management team, the general reaction was to treat the whole exercise with the cynicism that had previously been attached to apparently similar rescue attempts for British Steel and British Leyland. ICL, it was argued, was just another inefficient and ultimately doomed British company that was being kept temporarily afloat by the taxpayers' money - according to many judgments made at the time, it would have been better to cut the nation's losses and to sell whatever could be saved from the wreck of ICL to a suitable foreign-owned competitor.
Just a few months after ICL's near-collapse, however, opinion is now turning the other way. Partly, this is a direct result of energy and enthusiasm with which ICL's new management team, led by Chairman Christopher Laidlaw and Managing Director Robb Wilmot, have set about both resolving the company's long-standing financial problems and developing an exciting new strategy for re-establishing its once formidable world lead in computer development. In addition, the now detectable change of opinion is also an indirect result of the new management's efforts, which have forced observers to look again at both the many opportunities facing ICL and the many hidden strengths that the new management can call upon in trying to take advantage of those opportunities. While nobody would deny that re-establishing ICL's prosperity and reputation will be far from easy, few are now making the simplistic lame duck comparisons with British Steel and British Leyland, and the real parallel, but on a larger scale, may be with ICL's once major shareholder Ferranti, which in a relatively short time turned itself around from near-bankruptcy to profitable viability, all as a result of a Government-inspired rescue operation of the kind now sustaining ICL.
To understand why the future of ICL can be considered to be at least potentially bright, it is first necessary to understand the background to its present position and to that of the computer industry generally. From the start, ICL has been essentially a mainframe-computer manufacturer, basing its operations on the supply of the equipment and services needed to support the large-scale data-processing efforts of mainly commercial customers. Because of this, the company tends perhaps to be relatively little known to those outside the mainframe-computing world, which, ever since its real emergence in the early 1960s, has tended to remain largely aloof from the other electronics-based industries. While ICL, like all the mainframe manufacturers, has naturally had to retain a strong interest in basic electronics developments, it is probably true to say that the real links between the computer industry and the mainstream electronics-engineering community have been based on minicomputers and, more recently, microprocessors, rather than mainframes. Consequently, in spite of ICL's prime importance as one of Britain's few world-scale companies operating at the forefront of the new electronics-based technologies, it is probably true that the ICL story - that of its origins, past performance and present position - is almost unknown to the majority of electrical and electronics engineers.
Despite its recent dissociation from the rest of the electronics-engineering industry, however, ICL can certainly claim to have its roots firmly at the industry's heart. In the pioneering days of computing the construction of electronic computers, which could well contain upwards of 5000 thermionic valves, was naturally seen as an essentially electronics-engineering task. Consequently, it was to the traditional electronics-equipment manufacturers that the early computer pioneers looked when the time came to commercialise their ideas - with the result that the list of early computer manufacturers sounds like a roll call of the traditional electrical- and electronics-engineering companies: Ferranti, English Electric, Marconi, Metropolitan Vickers, GEC, STC, EMI, Elliott Brothers, Smiths Industries and Plessey. Of the many companies active in the early British computer industry, and thus destined eventually to become part of ICL, only two were outside the traditional electronics-engineering industry. Of these two non-electronics companies, one was perhaps the most unlikely type of organisation to pioneer an important new line of computer development, baker and tea-shop chain J. Lyons & Co. Ltd.
This enterprising company simply decided that it needed a computer for use in its own offices, and thus began construction of its own Lyons Electronic Office (LEO), one of the most influential of all the early computers.
The other non-electronics computer pioneer was, however, a much more obvious contender in the race to commercialise the new technology. This was the British Tabulating Machine Co. (BTM), a leading producer of the electromechanical (Hollerith) accounting machines that then made up a very important part of the whole business-equipment market. Of all the early computer developments, it was the work of Prof. A.D. Booth at Birkbeck College, London that came nearest to the kind of technology with which BTM was familiar and that fitted most neatly into its existing product ranges. Although Booth's original ARC relay-based machine was in some ways less advanced than its now better-known electronic contemporaries, its all-electronic successor, the APE(R)C, was soon re-engineered into BTM's prototype Hollerith Electronic Computer (HEC), and later marketed as the BTM 1200 in 1954. This move was especially important, as BTM later amalgamated with its main accounting-machine rival Powers-Samas, and then went on to progressively absorb all the other British computer manufacturers to eventually become ICL.
From a modern perspective, it is easy to see that this progressive amalgamation of the many early computer manufacturers into a single market-oriented group was more or less inevitable in the face of the realities of the fast-developing computer industry. While the technology-oriented electronics companies could dominate the scene in the early days of computing, when hardware was all important and computers sold only in ones and twos to predominately research-oriented users, the gradual emergence of a large-scale market for business-oriented computers put the electronics companies in a quandry.
To service the rapidly growing new commercial-computing market, it soon became clear, would require the development of huge marketing and user-support operations that would not fit in well with the operating patterns of the traditional electronics manufacturers, and which in any case threatened to overwhelm the more conventional activities of the electronics companies. Success in the commercial-computer market was clearly going to depend more than anything on close contacts with user organisations, and it was in this form of contact that the accounting-machine companies had long since specialised. Even if all this was not obvious in itself, it could be seen clearly enough in the case of IBM, the US-based business-machine giant, which throughout the 1950s was building up an overwhelming dominance of the world computer market that it has ever since managed to retain.
Because of these factors, and because it was also clear that successful operation in the world computer market demanded financial resources beyond the reach of typically sized British companies, constructive amalgamation was clearly the order of the day. Actually, this had been apparent to the most far-sighted of the early computer pioneers even before the first computer was commercially marketed, and, through the then newly formed National Research Development Corporation (NRDC), they repeatedly called for closer co-operation and faster development. Such urging received very little response, and, as the NRDC noted in December 1949: both the electronics manufacturers and the punched-card (i.e. accounting) machine manufacturers, respectively, represented that they were individually in positions to tackle the problems of an electronic-computer development project as well as, for example, the International Business Machines Corporation (i.e.IBM) in the United States. It was pointed out to the punched-card machine manufacturers that, in the opinion of the Corporation, they had inadequate electronic staff and resources. . ., and it could have equally been pointed out that the electronics companies had inadequate marketing staff and resources. Faced with a refusal of the British companies either to co-operate or to expand to the necessary extent, the NRDC began taking its own initiatives by supporting specific computer projects, and without this official urging it is unlikely that the British computer industry would have ever developed to the extent that it eventually did.
In fact, the sheer pressures of the market eventually made amalgamation unavoidable even for those who had initially reacted against it, and in this it was the market-oriented BTM that led the way. By the late 1960s, the whole of the British computer industry had come together into just two main groups, International Computers and Tabulators (ICT), which had formed around BTM, and the smaller if perhaps more technology-oriented group that had developed around the computer division of English Electric. Once that this considerable degree of rationalisation had been achieved, few people within the industry regarded further amalgamation as likely or even desirable.
The late 1960s, however, was the time at which the great post-war pro-amalgamation trend, or merger madness as it was also to become known, was at its height. In particular, the Labour Government's new Ministry of Technology, under the leadership of the then Mr.Anthony Wedgwood Benn, was greatly in favour of mergers between high-technology British companies, particularly in areas in which they were failing to keep up with their US-based competitors. As US penetration of the British computer market had increased from some 10% in 1960 to a level of some 50%, there seemed to be, on paper at least, many good reasons for pushing ahead with a merger between ICT and English Electric to form a single British computer company.
In fact, both companies were strongly against the idea. ICT thought it was doing well enough with its highly successful 1900 range, of which it had already sold some 1000 models by 1968, and regarded English Electric's new System 4 as an expensive non-starter. On the other hand, English Electric regarded the 1900 series as rapidly becoming outdated, and saw its forthcoming System 4 range, in which it had already invested huge sums, as being the key to the future. Over strong Government pressure, all the more telling because the public service was then by far the largest customer for computers, failed to force the two companies to reach agreement, with English Electric refusing to take less than 40% of the new company and ICT refusing to give it more than 25%. The result was a stalemate, at least until an unexpected intervention came from a most unlikely quarter.
The source of this unexpected intervention was the then enfant terrible of the electronics industry, Plessey. Eager to expand its operations further, and believing that ICT's relatively poor profit record would make it an easy acquisition, Plessey put in a surprise bid. At the time, the idea of Plessey controlling its larger rival in the national computer market was anathema to English Electric, and it immediately re-opened negotiations with ICT. Eventually, a compromise was reached in which the two companies would come together and allow in minority holdings from both Plessey and the Government, but the final terms turned out to be a very nasty shock to English Electric, which had suddenly hit severe problems with its System 4 development that seemed likely to demand at least an extra £10 million to put right. English Electric in fact received only 18% of the new company compared to the 53.5% that went to the shareholders of ICT, the 10.5% that went to the Government in exchange for a £17 million cash injection, and the 18% that went to Plessey for £17 million in cash.
This holding was not enough for Plessey, however, which suddenly saw a way to make itself both the largest single shareholder in the new national computer company and the owner of one of the giants of the electrical industry: it would simply take over the whole of English Electric. In the panic following Plessey's cheeky bid, English Electric finally came to terms with GEC, which had recently absorbed AEI and which now also absorbed English Electric to give Britain the single huge electrical and electronics conglomerate it has today. Plessey's original shareholding in ICL, for that was what the new computer company came to be called, was kept at its initial 18% level until relatively recently.
The net result of all this surprising manoeuvring, apart from the creation of the modern GEC, was the fulfillment of the Government's dream of a single flag carrier British computer company, which would be able to take on IBM at its own game. At the time, however, reaction to the merger was often far from favourable, and this was reinforced by the general unpopularity among commentators at the time of the strong 'Buy British' policy that went with it, according to which the Government strongly urged all public-sector users to buy only ICL equipment. While the defenders of this policy pointed out that it was much more liberal in practice than similar policies being exercised elsewhere, notably in the USA itself, the very existence of the policy led to ICL acquiring some of the unpopularity that the Buy British policy created as a result of the peculiar British aversion to all government involvement in commercial affairs. Besides this, however, there was also a feeling that the ICL merger had in fact achieved nothing, as it was widely believed that ICT could have carried on anyway as Britain's flag carrier company, with English Electric either staying on as a small-scale supplier or, in the view of many, simply pulling out of computer manufacture as so many other major electrical companies around the world were eventually to do. Indeed, it was also thought that, by loading down ICT with English Electric's somewhat dubious product range, all that had been achieved was the hobbling of the British computer industry. Whatever the truth of all these speculations, one thing is probably certain: it would have been a lot better for all concerned if the merger had taken place when the need for it first became apparent, perhaps 10 and at least 5 years' before it finally came about.
Nevertheless, the newly formed ICL was, on paper at least, now in a fairly strong position: indeed, the new company was able to justifiably boast of its role as the largest computer manufacturer outside the USA, and in many ways was able to put up a better fight against US domination than any of its other non-American national counterparts. Within its overall image as a single national computer company, however, was concealed a potentially fatal flaw: as in the case of a number of other mergers made at around the same time, simply bringing constituent companies into a single financial grouping was not enough to persuade them to work together as a unit.
In the case of ICL, the many different traditions contained within its operational structure made it difficult for some time to give the new group a real sense of identity and an overall sense of motivation, with the result that a certain tendency to what the new MD Robb Wilmot has called internal parochialism has been maintained right up to this day. Moreover, in ICL's case, the differences between its constituents were embodied and perpetuated in a particularly concrete form as a result of the company's sudden dependence em two essentially incompatible computer ranges, the 1900 and System 4. What to do about the two ranges was the most difficult question facing ICL at its formation, and indeed the consequences of this early dilemma still haunt the company today.
In considering ICL's dilemma of the two ranges, however, it is not enough to take into account the ICL machines themselves, for inevitably in the commercial computer world the real issues are concerned not so much with absolutes but with relativities, the relativities being in terms of comparison with IBM. In particular, the success of all other mainframe computer ranges to date must inevitably be measured in relation to that definitive range of machines, the IBM Series 360 and its closely similar follow-up, the Series 370. These computers, first introduced in 1964, transformed what had already been evident as a growing dominance of the world computer market by IBM into an established fact, resulting in a situation in which, by the 1970s, two out of every three computers in the world were made by IBM, and the giant company's computer sales were greater than its five closest rivals put together.
Much of this success is clearly due to IBM's unrivalled excellence in marketing and user support, which has consistently outweighed what has often been something of a weakness of the company in exploiting new technology. In the Series 360, however, IBM surprised the whole computer world with a range of computers so far in advance of nearly all those developed up to that date that since then the 360/370 range has essentially defined the terms on which all its rivals must compete. This is still as true today as it was in the mid-1960s, and indeed one of the remarkable facts of the computer industry today is how many companies are still finding it necessary to adopt an 'if you can't beat them, join them' policy by developing new products based on direct compatibility with the 360/370 designs.
Apart from the many architectural features of the Series 360 that contributed to its enormous success, however, there was another advantage to the design that in the end proved commercially decisive. This was that the Series 360 was a true range of computers, with models covering the whole spectrum of then available computing power from small-business systems to large-scale scientific number crunchers. The key aspect of this complete range of models is that they were essentially software compatible, i.e. in general a program written for one model in the range could in principle be run on any other model. In practice, this compatibility was not complete, but to a significant degree the whole range was upwards-compatible, i.e. a program written for a smaller model could always be transferred to a larger machine in the range, and was also planned to be forwards-compatible, i.e. users could write software for existing machines with an assurance that it could be transferred to newer models when they were introduced.
This concept of the software-compatible range proved to be the key to the real take off of the commercial computer market, which expanded rapidly in the years following the release of the 360. In Britain, for example, there were only 220 computers in use in 1960 (90% British made} but 6500 in 1970 (50% British made} and the 360 not only counted for a great deal of the growth in the overall number of computers but also contributed much more to the fast-growing penetration of the British market by US-designed products.
To appreciate the enormous advantages of the software-compatible range, it is necessary to consider the change that came over the computer market in the mid-1960s. In the 1950s computers were often built on a more or less once-off basis, popular models selling only in tens and the most popular selling only in the low hundreds. When a new technological development was introduced, or a new market demand satisfied, by a new model the new machines would have a unique design that made them essentially software incompatible with its predecessors. Initially, when software was very simple and the cost of its production was in any case dwarfed by the very high cost of electronics hardware, this lack of compatibility did not matter.
As software gradually became more complex, however, and the introduction of integrated circuits began to make hardware costs fall drastically, the investment that manufacturers and users were making in producing software began to dominate the overall cost of computing. This meant, in particular, that users found it very expensive to trade up to an incompatible larger machine when their computer requirements (as always) outgrew their existing model, and also that they were increasingly reluctant to switch to incompatible new models when they were introduced. The natural effect of this trend was to restrict the overall growth of computer usage, and it was this restriction that the software-compatible range very effectively removed. The 360 and its follow-up models sold not in tens and even hundreds, but in thousands and eventually tens of thousands.
These advantages of the series 360 have naturally become part of computer history, and IBM's enterprise in introducing them has also naturally been much commended. What has received much less attention, however, is the fact that very many of the same innovations were made at almost exactly the same time by the Ferranti/ICT design teams who began work on the 1900 series in 1961, leading to its introduction in 1964. Although using a typically British approach, for example, in having a 24 bit word divided into 6-bit bytes, the internal architecture of the 1900 was almost as modern in mid-1960s terms as that of the 360. Most important of all, the 1900 series was also a true software-compatible range, offering in fact a greater degree of upwards-compatibility and forwards-compatibility than the 360. These features meant that the 1900 range was able to compete very successfully, on a British scale at least, with the 360, and as well as over a thousand 1900s being sold before 1968, new models in the range went on being sold well into the mid-1970s.
With its 1900 series, as inherited from ICT, ICL thus found itself with a highly developed own-design range that could at the time at least hold its own with the 360. With its System 4 range, inherited from English Electric, however, ICL found itself with something quite different. To begin with, while the 1900 had sold 1000 models, the System 4 had sold very few, and, in spite of a substantial order book, the need for further development to iron out a number of design problems meant that the economics of the System 4 project began to look rather precarious.
In terms of its ability to compete with the 360, however, System 4 gave ICL, according to one view at least, a potentially enormous advantage for, essentially, the System 4 was a 360, or at least was a machine designed to be almost completely software-compatible with it. The System 4 range was, in fact, one of a large number of 360-compatible ranges introduced in the late 1960s by manufacturers who had become quite convinced that the advantages of the IBM architecture were such that it was essential not just to compete with it but to copy it. The actual System 4 design was based, with substantial modifications, on one of the most successful of the various IBM-compatible ranges, the RCA Spectra.
Finding itself with these two very different ranges, ICL was thus left in a quandary over what to do with them particularly as it seemed likely that its limited resources would not long enable it to support both. One line of argument within the company, naturally supported by the ex-ICT personnel, was that the successful 1900 range should be further developed, with the System 4 being quietly dropped. The opposing line of argument, naturally supported by the ex-English Electric personnel, was that the new IBM-compatible System 4 should become ICL's main product line, giving ICL a potentially very lucrative entree into the rapidly growing IBM-compatible market, and that the relatively dated 1900 series should be quietly phased out.
The pro-System 4 argument received some support from those who were convinced that the future of computing lay in interactive or online operation, which neither the 1900 or original 360 had really been designed to support, but to which the later design of the System 4 was specifically geared. This special advantage of the System 4 in online working, which in fact accounted for most of the extra design effort put into the System 4 development by English Electric, was certainly an important factor, but was to a considerable extent outweighed by the difficulties then being encountered in getting System 4 to work at all. In the end, the long-term attractions of System 4 received a devastating blow when subsequent aggressive price-cutting by IBM led to a drastic shake-out of the IBM-compatible manufacturers, which even included the complete departure of RCA from the computing scene. Within ICL, this created another motive, apart from the perhaps natural not-invented-here reaction to copying an IBM design, for staying well clear of IBM compatibility, both then and in the future.
The final decision on what to do about its two ranges was eventually reached by ICL after a long process of internal discussion. In the end, the decision taken was certainly bold, and some would say foolhardy: both existing ranges would be replaced as soon as possible by a New Range family based on the most advanced computing concepts then developed. In the meantime, the 1900 would remain as ICL's main range, with System 4 being supplied instead whenever a great deal of online working was envisaged.
Implicit in this New Range decision, everyone realised, were potentially enormous risks. For one thing, it meant that ICL would have to support, out of its limited overall resources, not only the development of a whole new range of especially advanced machines but also the simultaneous marketing of two existing ranges of incompatible machines. Secondly, by seeking to base its new range on the most advanced techniques available, ICL would be running all the risks of development setbacks and implementation teething troubles, at a time when its competitors were offering well-tried products and its users were demanding ever-more reliable systems. Thirdly, by moving to a non-compatible new range, ICL would be breaking the by then golden rule created by its own development of the software-compatible-range concept, namely of always maintaining forwards-compatibility and thus keeping users safely locked in to a growth path based on one's own machines. In the end, most of these potential risks were to turn out badly for ICL.
To begin with, the start of development of ICL's New Range (later to become known as the 2900 series) coincided with a new drive in the industry towards minicomputers and small-business systems. While minicomputer-like systems had been around for some time, they had up to then been more or less exclusively used in specialist engineering applications, but at the time the introduction of integrated-circuit construction was beginning to make small-scale machines potentially suitable for a huge range of new applications, both in engineering and in business. The actual widespread use of small systems in these new applications was to be delayed by the normal conservatism of computer users until the mid/late 1970s, but in anticipation of the coming boom, nearly all the existing computer manufacturers were working on new small-system designs, and a number of new minicomputer companies had already sprung up.
In Britain, the first of these new minicomputer companies had been Computer Technology Ltd. (CTL) formed by Ian Barron (now of Inmos) and which for a time after 1968 was Britain's only computer manufacturer other than ICL. After CTL, Ferranti and GEC were also to re-enter the computer market with new minicomputer ranges, as were companies like Digico, Redifon (Rediffusion), and the Computer Marketing Group (CMG). ICL, however, was by this time completely committed to its ambitious mainframe developments, and could spare no effort for developing small systems. Indeed, when the need arose from a slave minicomputer to front end its mainframes in online applications, it had to turn to CTL.
This lack of any ICL presence in the emerging small-systems market could well have continued almost up to date, had not the company benefited from a colossal piece of luck in the mid-1970s. This came about because of the financial difficulties of the US-based Singer Corporation, which as a result had to sell off several of its subsidiaries, including its Singer Business Machines operation, based in Stevenage, England. Singer had entered the small-business systems market very early (970) with a highly innovative system, System 10, which differed from its conventional minicomputer contemporaries by being geared to user friendly multi-user operation in a wide variety of business-oriented applications. System 10, and its small 1500 counterparts, had proved extremely popular and had achieved international sales in the several thousands by 1976, the year of Singer's withdrawal from the computer market.
Luckily for ICL, it was then able to step in and take over the whole Singer operation, including its US factories, for what was by all accounts a very reasonable sum, giving it a presence in the small-business market, which has grown, almost without ICL willing it, to account for up to one-third of ICL's turnover. Nevertheless, because of ICL's continuing need to get its new 2900 mainframe range established, as well as a certain degree of entrenched mainframe mentality within its management, the company has devoted very little of its development resources to its small systems. The result has been that its small systems operation has never made a real impact on the market, and that its small systems themselves have progressively become dated, with the System 10, for example, still being based on its original 1970 circuit technology, 11 years after its release.
As well as pre-empting nearly all of its development resources for years to come, the 2900 New Range decision also opened ICL to all the development troubles that can accompany an innovative new product range. Again, the risk taken in the New Range decision turned out badly, for the original 2900 models released in late 1974 were quickly beset with all sorts of teething troubles. The worst troubles tended to occur with the top-end models of the range, the 2980 number crunchers, which soon began to have a reputation of working at no more than a fraction of their stated performance. Other troubles were also rife, including problems with the new VME operating systems. Troubles such as these not only hit ICL hard in terms of damage to its reputation and potentially lost orders, but they also required the commitment of yet more of its development resources in order to get them solved, reducing yet further the resources that could be devoted to developments in new areas.
Worst of all, however, was the New Range decision's consequence that software compatibility with the preceding 1900 (and for that matter System 4) machines was to be broken. The result of this was that, in general, 1900 users transferring to a 2900 had to re-write all their software, a requirement that, in spite of a number of conversion aids provided by ICL, was bound to put a very heavy and expensive load on their programming staff. As an example of the high costs of software conversion generally, just converting all the software regularly run on a typically mainframe installation from one operating system to another may well involve around 25 man-years of programming effort, and, naturally, converting from one whole type of machine to a very different one is likely to be even more onerous.
As far as a computer manufacturer is concerned, however, losing forwards compatibility by requiring such conversion has a much worse effect than just placing a large burden on its users, for it opens up the possibility that they will no longer stay locked into the manufacturer's range: if a user has to expensively convert all his software from an ICL 1900 to an ICL 2900, why should he not at least consider at the same time an alternative change to, say, a new IBM machine? As a result of breaking in this way the bonds that locked in its users to its own machines, ICL at one stage lost a number of important users to other manufacturers, a heavy blow for any mainframe company to suffer.
Actually, in spite of the fact that the 2900 machines are based on a completely different architecture to the 1900 range, there is no reason, in theory at least why they should not be made at least in one sense software compatible with their older counterparts. This is possible through the use of microcode-based emulation techniques, which allow one computer to perform effectively as if it were another, thus making it capable of running the other machine's software. Such emulation was certainly known to be possible within the 2900 range, and indeed the 2900's architecture makes it a relatively efficient emulator compared to many other machines, but emulation is in any case only possible on those models which have microcoded processors, which until recently ruled it out for the large non-microded machines at the top of the range. The desirability of providing such emulation facilities on the 2900 was widely discussed within ICL during the new range's development and indeed it was decided that the 2903/4 mini mainframe models at the very bottom of the range were subject to special market forces which made emulation of the 1900 machines desirable - these machines and their new ME29 successor, have in fact proved much more popular than originally envisaged and currently account for an unexpectedly high 30% or so of ICL's turnover.
Nevertheless, for the other machines in the 2900 range, however, the idea of providing emulation facilities was eventually rejected. The reason for this was that at the time it was seen as more important to encourage users to exploit the advantages of the 2900's own advanced architecture than to allow them to operate their new 2900 machines comfortably but inefficiently as 1900 emulators. From ICL's point of view this decision was natural enough, for, having gone to all the trouble of producing a new advanced-architecture range that it was convinced could compete very effectively on its own merits, it was not keen to effectively hobble its new machines by having them used just like their predecessors with their 15-year-old architecture. From the users' point of view, however, the short-term continuity that could have been made possible by the early provision of emulation facilities proved to be, rightly or wrongly, more important to their conservative outlook than exploiting the new architecture's long-term advantages, and the fact that ICL was known to be deliberately withholding the DME emulation facilities it had already developed was the source of much friction. Only after a number of important users had gone over to other manufacturers, however, was the DME emulation facility actually released, but the damage had already been done. Even today, a large proportion of 2900 users still use DME to run their 1900 software on their new 2900 machines, in spite of continuing ICL attempts to persuade them to convert.
As a result of all these considerations, ICL can be seen to have suffered greatly in the mid/late 1970s as a result of its original New Range decision. Nevertheless, the 2900 range has now established itself as very much a technical success, even though short-term commercial factors have so far tended to counteract the many technical advantages of its new-generation architecture. There are growing signs, however, that the commercial pressures that have tended to fossilise mainframe computing on the lines of the old-generation architectures such as the 360/370, and to penalise attempts by pioneers such as ICL to introduce advanced new architectures, may at last be breaking down, producing what new ICL MD Robb Wilmot terms a discontinuity in computing practice. The result is that even IBM is showing signs of adopting a new architecture for its future mainframes, although it is sure to provide DME-like emulation facilities for its existing users, and in this new environment the 2900 may well come into its own as commercially the most established of a whole range of new mainframe architectures that may be progressively introduced throughout the 1980s. Certainly, ICL expects working 2900 systems to be around long after other older-architecture ranges are recognised as being commercially as well as technically obsolete.
Until that recognition comes about, however, ICL will continue to be under pressure from some quarters to finally switch to an IBM-compatible architecture. In a sense, this is a resurgence of the old 1900/System 4 argument that bedevilled its early days, for although the original 360-compatible market, of which System 4 was part, eventually collapsed in some ignominy, since then a thriving new 370-compatible market has arisen. So great are the commercial rewards now to be made in this new market that it seems as if almost every mainframe manufacturer is introducing new IBM-compatible products. ICL, which has never suffered any shortage of commentators ready to tell it how to run its business, is now being urged by many to cut its losses on its independently minded 2900 development and to develop 370-compatible versions of its machines so as to exploit the booming IBM market.
This feeling that ICL should turn towards an IBM-compatible strategy apparently received some confirmation recently when it was announced that the company was to market the large 370-compatible mainframes in the Fujitsu Facom 380 range. However, ICL insists that this arrangement is a purely commercial one, and that its 2900 developments will proceed even faster in future than in the past: while it is willing to sell IBM-compatible machines, it wants to do so merely as special-purpose components of large ICL-based networks. Still the possibility of ICL starting to sell its 2900 machines with DME-370 emulation facilities, has been widely discussed, particularly as ICL is known to have already developed such facilities for research purposes. The arguments over the proposal, however, closely parallel to those over the release of the original DME-1900 facilities: is it best to go for short-term gains or to push ahead with exploiting the long-term advantages of the advanced 2900 architecture?
Apart from the still continuing arguments over IBM compatibility, however, the last few years have seen a growing general belief in ICL's ability to make its bold 2900 development finally payoff. Indeed, throughout the 1970s the company had, in spite of all its difficulties, managed to grow at a greater rate than almost all of its rivals, including IBM, and by the end of the 1970s things were beginning to look very much better for the company in all its spheres of operation. Unfortunately for ICL, however, the current recession, which has naturally hit the mainframe market hard, producing a zero net growth in sales as opposed to the consistent 20% growth experienced for a decade or so, caught the company unprepared for a financial squeeze. As is now clear, ICL's cost-control procedures proved quite unable to keep costs from rising steadily as sales revenues stagnated, and the inevitable result was the financial crisis that hit the company early this year.
Naturally, one of the first moves of ICL's new MD Robb Wilmot was to put in hand a series of tough measures to overcome the company's cost-control problems, measures including new production controls that will cut back inventories by £40 million and a reduction in manning levels that will cut the payroll by some ££60 million a year. As well as these short-term cost-control measures, however, Wilmot has also announced extremely important changes to ICL's product strategy, and it is these that give rise to the now general, if inevitably still guarded, optimism about the company's future.
In just 6 months, Wilmot has clearly managed to get a firm grasp on the company, and to formulate new policies that make the most of its many real advantages - its now established advanced mainframe range, its growing range of new small systems, its strong background in computer networking and its highly regarded worldwide marketing operation. At the same time, he has clearly put an end to some of the internal organisation weaknesses, often based on what he has condemned as divisional parochialism, which have led to inefficiences in the past - for example ICL's support of no less than 10 operating systems, which Wilmot has now ordered to be progressively combined into a single VME 2900 system. Above all, he seems to have instilled a new faith in ICL's future not only into its formerly somewhat demoralised staff, but also into the management of some of the most successful high-technology companies in the world, which he has persuaded to join him in important new collaborative ventures. With his far-reaching strategy to focus the company's efforts on its new Networked Product Line, Wilmot has now given ICL's business plans a conviction they have for long lacked. The ICL story may, in fact, be about to enter an exciting new chapter.
Under new management...
Within only 6 months of taking up his post as ICL's new MD, Robb Wilmot, electronics engineer and former head of Texas Instruments' British-based calculator operation, has not only introduced tough measures to cure ICL's short-term financial crisis, but has also overseen the announcement of a number of extremely significant new developments for the company. These new developments, which include not only innovative new product ranges but also Important new collaborative agreements with leading foreign-based companies, are all based on Wilmot's conception of ICL's future role in the computer industry, as embodied in his far-reaching new plans for an all-embracing ICL Networked Product Line.
The key to Wilmot's ideas about the future of the computer industry, and of ICL's role in it, seems to be the concept of what he identifies as approaching discontinuities in computer technology, computer marketing and, above all, computer usage. In terms of technology, these discontinuities include, for instance, the switch from mainframe construction based on SSI/MSI ECL devices to a new VLSI CMOS approach, which will have tremendous implications for mainframe design. In terms of marketing, the discontinuities include the shift from the traditional computer-sales approach to that of pop in computer shops catering for small-scale first-time users. In terms of computer usage, the discontinuities include not only the much publicised switch to personal computers of all sorts, but also the less remarked switch in mainframe operation from 24-hour 3-shift working to an ordinary 8-hour single-shift day. All these discontinuities, Wilmot points out, create opportunities for a company ready to exploit them.
As Wilmot says of the coming developments in the computer industry, the opportunity is potentially infinite - but the challenge is how to manage these infinite opportunities effectively. To those who have been misled by the extraordinary energy of Wilmot's short tenure at ICL, which has already resulted in wave after wave of product announcements, and which may have given some the impression that he has simply been jumping onto every passing bandwagon, Wilmot also makes clear his recognition that the danger is trying to do everything and not doing anything properly. What is needed, he says, is a clear focus. The focus for ICL is that we will drive our product activity in the direction of a Networked Product Line. Behind Wilmot's emphasis on the Networked Product Line is the recognition that, no matter how great are the benefits of installing new stand-alone computer and electronic office equipment, much greater benefits will accrue from operating such equipment together in a network. Once you join otherwise freestanding computers together with a length of wire and apply software, he points out, the benefits multiply. No major computer user should allow proliferation of computers - micros or minis or word processors without a networking strategy to yield these additional benefits in future. Indeed, even small businesses buying their first computer have a need to base their growth on future networked systems as a fundamental requirement. So basic to his plans is this network strategy that he makes it clear that products which do not fit into the Networked Product Line, even personal computers, will not in general be marketed by ICL.
In terms of the actual implementation of the new computer networks, Wilmot has already done much to make the Networked Product Line a reality. To begin with, ICL's traditional mainframes can already participate in advanced wide-area networks (WANs) as a result of the company's development of its comprehensive Information Processing Architecture (IPA). All ICL products will in future be able to interact via IPA, as well as, in many cases, via the IBM networking scheme SNA - the latter being supported, according to Wilmot, in recognition of the fact that many IBM mainframe users will, we believe, develop an appetite for working with ICL networks. As well as the traditional WANs, ICL will in future be supporting a range of local-area networks (LANs), which will interconnect equipment on the same site. Already announced is ICL's new 1 Mbit/s Microlan, which is being used to interconnect its new DRS small-business systems (IEE News, Nov/Dec. 1981, p.17), and this will be followed by a coaxial-based 10 Mbit/s high-speed LAN, and eventually by a much higher-speed Macrolan. For those users who prefer the PABX route to local interconnection, however, Wilmot has also provided ICL with the marketing rights to Mitel Corp's new high-technology SX2000 PABX. On any given site, PABX and LAN interconnections will be capable of being mixed as required.
To support such networks, Wilmot has already introduced two new small-system product lines that, although obviously under development before his advent, have had their fortunes greatly boosted by his new approach. Within ICL, such small-system products as DRS and the new System 25 (successor to the ever popular but now very dated Singer System 10) have received little encouragement by way of development resources, some 80% of which has in the past been devoted to the company's generally much less profitable mainframe range. Under Wilmot's direction this consequence of the company's traditional emphasis on purely mainframe developments will certainly be reversed: ICL is going to be big in small systems, as he simply puts it.
To make itself big in small systems, however, ICL is clearly going to have to increase considerably its investment in small system development. Wilmot acknowledges this, and forecasts that it will be necessary to ramp up small-system development funding to perhaps 12-15% of revenue, or even more. Where will the money come from? From the development funding now devoted to the company's mainframe range, which Wilmot has already partially axed by scrapping all the less-profitable models at the top of the 2900 range. Does this mean that ICL will be pulling out of the mainframe market? Certainly not, according to Wilmot, who says: ICL is going to be big in big systems too.
To manage this, while greatly reducing mainframe development costs, Wilmot has quickly set about rationalislng the range. Noting that the company's new 2966 mid-range design (the best mainframe ICL has ever produced) is considerably less costly to build than the older designs, a feature it shares with the new ME29 mini-mainframe, Wilmot has scrapped all the 2900 designs except these two and the entry-level 2946. While improving further the 2966 itself (e.g. by moving up to 64 kbit RAMs), Wilmot has produced lower-power (2958) and higher power (2988) versions and has also introduced more dual-processor versions of the same design. The result Is that he can now cover an expanded range of performance with just two basic designs instead of the previous six, and yet get greater per-unit profitability as well.
In the medium term, moreover, Wilmot is planning yet more powerful 2900 models by means of his new agreement with the Japanese giant Fujitsu. Having chosen Fujitsu because he believes It has the most advanced mainframe-oriented semiconductor technology in the world, Wilmot has persuaded the Japanese company to give ICl access to Its chip technology and its CAD systems. The result will be a new ECL VlSI-based 2900 model codenamed Espriel, which will be released around 1984, and which, in comparlslon with the 4 million instructions/s (MIPS) of the dual 2988 (the largest mainframe ICL has ever produced), will have a basic performance of 7 MIPS, with a dual version at 12 MIPS, and a quad version of 20 MIPS.
Also around 1984, ICL will be releasing a new Distributed Mainframe (DM1) which will use Fujitsu's CMOS VlSI technology to make it a true office-system mainframe. What this means is that the new model, although operating at a respectable 0.8 MIPS (1.4 MIPS dual, 2 MIPS quad), will require no air conditioning, special power supplies, or dedicated operating staff - a mainframe operating just like a modern minicomputer, in fact. Moreover, by 1988 the DM range will include 2 MIPS basic models (4 MIPS dual, 8 MIPS quad), while the Espriel will be 15 MIPS basic (25 MIPS dual, 50 MIPS quad). In comparison, the most powerful general-purpose mainframes available today operate at around 8-10 MIPS.
In developing these plans for the 2900 range, Wilmot has sought to get the maximum advantage from the advanced nature of its basic architecture. The ease with which both dual and quad multiprocessor models can be built up from the basic machines, for example, owes much to the 2900's internally distributed design. This feature is also being built upon to give a new nodal machine configuration that will lead to a form of practically 'non-stop' operation more or less unknown in the mainframe world the new 2900 models will in fact be able to operate continuously without manual intervention for mean times between system outages of some 75000 hours.
Clearly, Wilmot's mainframe plans depend heavily on the link with Fujitsu, but how has he persuaded the Japanese company to give him access to its much prized new technology? Partly, by his use of ICL's greatest asset, its skill in marketing computer products, and particularly mainframes, which require very detailed user support. As Wilmot points out, during practically the whole of the 1970s ICL grew faster than almost all its competitors, largely as a result of its highly skilled marketing teams, which alone include over 2000 experienced software engineers.
Although because of the recent fall-off In sales this marketing team is now under-employed, which has been one problem contributing to the cost overruns that have led to ICL's financial downfall, Wilmot has effectively turned the problem on its head by pointing out that the situation is more usefully viewed as ICL having some $500 million per year of spare marketing capacity. Having seen the situation In this light, he has thus gone all out to exploit the advantage it provides.
As Wilmot points out: This "instant" marketing capacity is unique in the industry and represents an extremely valuable asset to ICL, particularly at this time when the computer Industry is poised at the threshold of the VLSI era and networked computing is about to pervade the office, the factory, engineering and research, warehousing, retailing and the financial world.
While many companies are now beginning to produce what he calls unbelievable cost-performance VLSI-based products, he points out that few can afford the expense and time needed to build up a proper worldwide marketing operation, particularly as success in the VLSI-based market is likely to go to those who build up their market shares most rapidly. With its spare marketing capacity, Wilmot believes, ICL is ideally placed to help others short-circuit the marketing problem, and it is this marketing service that he has been using to entice into collaborative agreements not only Fujitsu but also companies like Three Rivers Computer Corp., the originator of the new ICL PERQ system.
As well as ICL's conventional marketing operation, however, Wilmot is also keen to exploit new ways of reaching customers, particularly with his new small systems. While in the past, ICL, like other traditional mainframe companies, has been extremely reluctant to allow any third party to become even marginally involved in the marketing of its products, Wilmot is now keen to develop all the third-party channels to potential customers that he can find. As a result, ICL is not only opening up a world-wide chain of 'Computer Point retail outlets for its products, but Is also forming a parallel Trade-Point organisation geared to supporting third-party sales. ICL will also allow, for the first time, OEMs to incorporate ICL products into their own systems, which may or may not subsequently bear the ICL label.
According to ICL, dealers have already been inundating the company with requests to market its new systems such as DRS - not a bad turn-around for a company that, just before Wilmot came along 6 months ago, was thought likely to be sold off to any willing bldder.
1 LAVINGTON, S.H.: 'The early days of British computers', Electr. & Power, 1978, 24 p827 and 1979, 25. p4O
2 LAVINGTON, S.H.: 'Early British computers' (Manchester University Press, 1980
3 BUCKLE, J.K.: 'The ICL 2900 Series' (Macmillan, 1978)
ICL in the European computer market
* includes 30% from process-control equipment
+ anomalies due to currency fluctuations
Source: Logica Ltd., as published in Datamation, Sept. 1981
European computer market rankings, 1980. Like the overall world market, the European computer market is overwhelmingly dominated by IBM, but partly as a result of national buying policies, the 4 main European flag carriers at least manage to outrank the other US-based computer giants.
Of the flag carrier companies, Siemens has on the face of it the largest share of the market, but in practice this is offset by the fact that around a third of its computer sales actually consist of process-control systems in which the other more traditional companies have never interested themselves - in any case, a full 80% of Siemens sales are made in Germany and its computer operation has not made a profit in 20 years'.
Much more in the mainstream of the traditional computer market are Britain's ICL and France's Cii-Honeywell-Bull, which have jostled for years for the title of Europe's largest indigenous computer company. ICL, in fact, has generally kept the lead, although with the present setback of its financial difficulties, and with its French rival benefiting greatly from a still booming home market, the race is still undecided.
The French company has in any case been, in ICL's view at least, somewhat restricted in its ability to claim the European title because of its remaining 47% ownership by US-based Honeywell Information Systems, but this large foreign share holding is in fact likely to end soon as a result of the Mitterand nationalisation programme. Also, it must be said, that being the largest European-based company is no longer quite the accomplishment it once was: the days when ICL could use its European leadership to proclaim itself the largest non-American computer manufacturer in the world, are now at an end, thanks to the startlingly rapid growth of Japanese companies such as its new collaborator Fujitsu.
ICL does not necessarily accept the complete accuracy of this article, but endorses it as an excellent review of ICL's history to the present day.