A variety of bus standards now exists to describe the ways in which computer boards and systems communicate, covering 8-, 16-, and 32-bit wide interactions, though they are usually orientated towards a specific processors or applications. Fututrebus+ is the result of a ten-year effort to go beyond 32 bits and define a non-proprietary, scalable standard – up to 256 bits – which is independent of any particular technology. Indeed by the end of 1990 a set of Futurebus+ specifications fully ratified by the IEEE should be available to the industry. A report by Oxford-based Elsevier Advanced Technology is bullish about the prospects for Futurebus+ for two reasons – its high performance – achieved through multi-processing and a backplane throughput speed reckoned to go from 80Mb to 3,000Mb per second – and its open systems features, claimed to be far in advance of any competing bus system. Furthermore, Futurebus+ proponents argue, the specification is futureproof, and can grow to accommodate changes. They claim for example, that a workstation built with Futurebus+ technology in 1993 could potentially be upgraded with the latest processor in say 1997, without having to scrap large parts of the system. Industry-specific subsets – or profiles of Futurebus+ can be created covering different hardware and operating environments so that various implementations will conform to the standard. Profiles for DEC – which is the most advanced of major vendors with Futurebus+ development – VMEbus and workstation manufacturers, as well as for the military and telecommunications industries are now being created. However widespread adoption of Futurebus+ is being held back by the the lack of protocol and management chips, says the report, a shortfall which will it does not expect to be made up until the end of next year.
Interface elements expensive
Futurebus+ interface elements are also expensive, require a lot of board space and there is a lack of multi-processing Unix software available. Nevertheless the report confidently predicts that Futurebus+ will become the highest selling bus architecture by the end of the decade, when it will account for an estimated $2,300m of computer systems sales. The 50 or so companies and 15 universities currently working on Futurebus+ technology are estimated to have invested something in the region of $40m so far, which is expected to generate $200m of activity by 1992. Military and mainframe applications of Futurebus+ technology will emerge by this time, says the report, followed by workstation and telecommunications implementations sometime later. Although the camps which support other bus standards such as VME and Multibus II – have declared support for Futurebus+, the report is highly sceptical of at least some of their intentions. Intel appears to be strongly promoting Multibus II as a Futurebus+ substitute, and has made a number of moves recently calculated to attack the 1990-1995 market for Futurebus+, it claims. Mainframes will lead the way in the use of commercial Futurebus+ technology – the report expects input-output controllers, concentrators and cluster servers that work with mainframes to be announced over the next year. In particular it reckons DEC Europe – as opposed to DEC US – will announce a 64-bit Futurebus+ implementation of a VAXcluster controller or similar to aid throughput between machines, said to be the first of five mainframe peripheral and networking projects that will come out of DEC UK and Ireland. IBM will of course be backing Micro Channel against any inroads from Futurebus+… The 60-page report costs UKP900.
