Every mainframe vendor has users still addicted to old machines that are no longer in production, and ICL is no exception to this rule. Over the past three years the company has been trying to ease the transition to its Series 39 range for ME 29 users, and ICL’s manager of mainframe marketing Derek Sayers says that these users are still moving across. However, many users seem to believe that a move to an equivalent-sized Unix machine would be more sensible than a move up to VME.
If this turns out to be a widespread belief then ICL may face the same problems from this user base as it has from System 25 users, who are now moving into the DRS6000 market rather than feeding into the Series 39 base as ICL must have hoped. Nevertheless, Sayers is adamant that VME is the best operating system in the world, faster and more flexible than IBM’s MVS. He foresees three future directions of development for VME: firstly, at the end of the year ICL will launch its Open System Control Centre, which is designed to sell the concept of ICL as a supplier of a co-operative client-server environment, embracing ICL’s Oslan networking components, its DRS6000 workstation and VME. The Centre will run the network using a knowledge-base kernel, which automates system management. Secondly, VME will move sideways by conforming to the X/Open Portability Guide 3, so that applications may be downsized from the mainframe. The development work for this has built on VNS, ICL’s Unix offering running under VME. Of this early stab at combining VME and Unix, Sayers says the functionality was perfect but the performance was not good, because Unix’s screen-based structure slows down the interface to the character-based VME. Furthermore, VNS didn’t benefit from CAFS or Series 39 security features. These problems will be rectified in the new XPG3 interface to VME. Thirdly, VME will be developed upwards to support hundreds of thousands of terminals, terabytes of data, and the number of jobs it can run – Essex will be a 360 MIPS system in two years. VNS did not take off and has been withdrawn from active marketing. As for Unix running native on Series 39, Sayers is sceptical (which will frustrate a lot of ME29 users), saying that there is no substantial market demand for such a machine. Amdahl, he commented drily, has the majority of its Unix mainframes running at AT&T where they are used to develop Unix. He joined in with the proprietary vendor chorus thattrills that open systems will always lag behind the technical advances of proprietary offerings. He added that it just doesn’t make sense to run Unix on the mainframe, since it can’t handle the necessary input-output throughput or offer secure data management.
By Katy Ring
As far as mainframe development goes ICL has other fish to fry aside from Unix. For example, it is undertaking an ESPRIT project with Siemens AG and Bull SA to develop large parallel processing systems. Sayers believes that ICL will bring a product to market within five years which will provide transaction processing performance 100 times faster than anything available to today’s commercial market. He says the product will be a back end parallel computing system running as a database engine, with relational type of access. Talk of mainframe development brings ICL shareholder Fujitsu to mind. Sayers admits that ICL already has a good cost-effective and intimate collaboration with Fujitsu – collaboration which has proved itself with the arrival of the Series 39 mainframe range and most recently the launch of the top-end SX models. However, Sayers stresses that the relationship is mutually beneficial, pointing out that before its partnership with ICL, Fujitsu had stuck to copy-cat designs of IBM silicon technology. But with SX, for example, ICL gave Fujitsu the expertise to design very small silicon pieces. According to Sayers, it cost UKP200m to bring SX to market, and without Fujitsu, he says, it may well have cost a further UKP100m, which would have eaten into ICL’s profitability, or meant the company delaying the mainframe launc
h for another year. Fujitsu is, of course, also helpful in providing peripherals and point of sale systems. Yet, because of ICL’s current intimacy with Fujitsu, Sayers had to say that in his opinion there were no additional benefits to be had from ICL becoming a 100%-owned Fujitsu subsidiary. As for the other frontrunner to take control of ICL, AT&T, Sayers smiled diplomatically and said that yes, there was a close relationship developing between the two companies because of their work for Unix International. Of course AT&T’s intentions vis-a-vis ICL remain undeclared, but as reported last month (CI No 1,432) it is sniffing around ICL’s user base by bidding for Kleinwort Benson’s leasing operation Park Leasing which serves 83% of ICL’s installed UK base.
Exchange Hire Ltd
Sayers was rather defensive about the whole issue saying that ICL controls its own user base via Exchange Hire Ltd, while Park merely controls the funding of Exchange Hire users. He claimed that all that AT&T would have access to is a list of user names and addresses. However, he then admitted that ICL does plan with Park what products will be brought to market and when, so that the leasing company will know what timescale it is working to, enabling it to swap customers over to the new technology at the right time. But, the concept of Exchange Hire is owned by ICL and if it felt that liberties were being taken it could fund the business via another channel. Sayers said that only a few ICL customers were effected by the Atlantic crash and that the company is talking to them to try and get their financial arrangements straight. He was not in a position to comment on the intimation that STC’s recent profit warning was a result of ICL research and development costs, but expressed surprise that the finger should be pointed at SX development. As for the sort of collaboration Sayers would like to see with possible partners or future owners, he felt that approaches would be welcome by companies that could help ICL bring its parallel processing project to market, or could help develop ICL’s client-server approach.