The way IBM’s top end mainframes must develop to dig it out of its financial mess IBM is in a bigger financial mess now than it has been for the last four years, particularly in the US. Understand this, and future product and software pricing trends within the IBM mainframe arena assume a degree of inexorable logic. This was the essence of the message delivered to Xephon IBM Futures delegates by Barry Graham of National Advanced Systems, who also managed to throw in a few tips on how to reap maximum financial gain from IBM’s discomfort. According to Graham, IBM has seen a $2,000m decline in US revenue growth since 1984. Add to this the soaring US dollar, and a disastrous first quarter in IBM UK, Germany, and France, and two courses of action become inevitable. One – IBM will launch a major new machine – Summit – shortly. Two – software prices for large system users, particularly those in Europe will rise. On the first score, Graham expects Summit to ship early next year – he’s out on a limb on that because most people don’t expect it to ship until 1991 – close on the heels of a 3090 speed-up. He believes the initial model will be a four processor system, based on a 27 MIPS uniprocessor, will halve cycle time to between 7.5nS to 9nS, and occupy about half the physical space of its predecessor. He also believes that it will offer sufficient storage to exploit both Extended Systems Architecture, and PR/SM, thanks to an initial 1Gb of main storage, and the use of second level cache – a buffer placed between the existing cache buffer, and the processor. Is Summit a good buy? On pricing, Graham points to a figure of $8m or UKP9m – double the cost of the 3090-200, when it was first introduced. Once Summit is announced, Graham argued, customers are forced into one of three corners. They either opt for Summit, which provides more MIPS than they need; buy an old system, which rapidly becomes obsolete, and needs replacing; or buy an additional mid-range system, thereby increasing their software costs. There is no answer to the question is Summit a good buy?, Graham concluded, simply an acknowledgement that an additional old system is the wrong choice, and the new one too expensive. In the meantime, he advises users upgrading to a second-hand 3090E to make sure that they can get to an S Model, and ensure that any 3090 processor discount also applies to memory storage upgrades. He also warns that because of IBM’s current financial difficulties, a major series of price reductions is unlikely to occur until just before Summit arrives; traditionally IBM has made dramatic prices cuts at the end of a product’s life cycle, because it can’t introduce a new machine at a higher price than the one it is intended to replace. Graham also believes that customers looking for a 4381 replacement – the so-called 4391 – will have to wait until Summit has been shipped. The key to the argument is his assertion that IBM is recycling Thermal Conduction modules, displaced from 3090 200 and 400 Base models by users upgrading to Es or Ss. According to Graham, these TCMs are rebuilt and used to create 3090-120S, 150S, and 170S machines, whose sales would be obliterated with the arrival of the 4391. Consequently, IBM will wait until sufficient sales of the low-end 3090s have been made to recoup 3090 investment, launch Summit, and then – and only then – produce the 4391. On the software pricing front, Graham expects that the AS/400 policy, where charges reflect the use or number of connected terminals, will gradually be applied to all IBM software products. He predicts that annual software budgets for large system users will increase by some 30% to 35% over the next few years, but also believes that price increases will come hand-in-hand with major improvements, such as the growth of dedicated artificial intelligence, supercomputing, database, and communications processors. He also expects the VM operating system to play a more prominent role in future IBM plans, but counsels patience: software changes take much longer than hardware ones to implemen
t, warned Graham, rounding off the session. – Sophie Hanscombe The vital need for network management Company directors, all too used to being reminded of the value of information technology in achieving competitive advantage, would do well to heed such advice where it concerns network management, according to Logica Plc consultant Richard Lavender. Many firms’ business competitiveness is now totally dependent on the facilities provided by their on-line terminal network, he warns. The demands of these network management facilities, constructed to monitor and report network activities, are growing with requirements such as round the clock operation. At the heart of IBM’s Network Management Architecture is a focal point consisting of NetView program products. But, he says, the notion of NetView can be misleading as there are products outside NetView that have a direct effect on the operation of IBM networks. One is NetView Performance Monitor: although a component of the focal point in Network Management Architecture, it is not integrated into NetView, and there is no need for NetView to be present in any host configuration. Systems Network Architecture has never been static, with many strategic changes since its conception in 1974. For instance, when PU type 2.1 architecture was defined, the central principle that there should always be LU-SSCP sessions present before any LU-LU communication was possible was broken. The new architecture allowed peer-to-peer communication personal computer-to-personal computer, or AS/400-to-personal computer, without the need for an SSCP; the architecture also relaxed the local addressing structures. Lavender predicts that IBM’s dominance will be challenged in many networking sectors, particularly in integrated services digital network, X400, X500, X25 and voice transmission; File Transfer, Access and Management and VTP; MAP and TOP; and a range of transmission technologies such as TRNs, 802 local area networks and packet switched data networks. How NetView will relate to the OSI/Communications Subsystems is unclear, but Lavender thinks that with the arrival of SAA-defined OSI components, NetView will continue to play a central role. This is because there are still SNA sessions present, as the only method that allows VTAM applications to communicate with the outside world is the SNA LU-LU session, and OSI/CS is a VTAM application. Looking to IBM’s plans for NetView, he sees an important omission in the ACF/VTAM application programming area relating to LU6.2 sessions. In ACF/VTAM Version 3 Release 2, there are facilities that allow high-level macro calls to manage and control LU6.2 sessions for basic conversations. Though most of the APPC verbs are represented, there is no simple method of providing a control interface to manage an APPC LU. For Lavender, NetView is an ideal vehicle for providing a control operator interface to APPC LU. This is the interface that includes sessions to be activated, session limit changes and the definition of new resources. But NetView is still missing an LU6.2 interface so little can be done either for PU type 2.1 nodes with independent LUs, or for APPN network nodes connected to SNA subarea nodes. The architecture of a type 2.1 node should be enhanced to include provisions for network management, and now NetView can exist in multiple address spaces under MVS, it should be possible to partition the OSI management and X500 directories. He concludes that IBM will definitely follow OSI Management Architecture, in an effort to ensure that its machines will be able to exploit the new networking capabilities brought about by OSI. – Andy Evagora
