This paper chronicles the involvement of the Science and Engineering Research Council's Engineering Board supported community and Prime Computers from 1976 to 1988.
Following some essential background information, the paper details how Prime machines were introduced to the UK Academic Engineers as part of a distributed Interactive Computing Facility and what special features and facilities were offered.
The paper ends with a brief but positive summary of what has been a major success story over the past 12 years.
The Science and Engineering Research Council (SERC), which was established in 1965, is a public body responsible for the support of basic and strategic research in pure and applied science (including engineering and computer science) outside the specialised areas of agriculture, medicine and the environment. Basic science is that which seeks to advance our understanding of the natural worlds for broadly defined and long-term ends. It provides the essential underpinning for the solution of technical problems and the new ideas which will be exploited for industrial innovation. Strategic research is more finely focused on the solution of urgent problems of national importance. Its funding comes from the Government's Department of Education and Science.
The Council promotes research in the universities and polytechnics by grants and studentships. It also supports university research by providing central facilities both nationally and internationally. Particular examples of its international activities are the telescopes in the Canaries and Hawaii, the support of the CERN Nuclear Physics Laboratory in Geneva and Institut Laue Langevin in Grenoble.
SERC breaks its activities into four main areas supported by separate Boards:
Its total annual budget is around £350M; it employs about 2700 staff and supports a further 6500 on grants in universities and polytechnics.
The Rutherford Appleton Laboratory (RAL) is the largest establishment of the Science and Engineering Research Council. It is situated half way between Oxford and Newbury, just off the A34, close to the Berkshire Downs.
The national support for research within SERC's remit is provided by central facilities at RAL and three other smaller establishments, at Daresbury, Edinburgh and Herstmonceux. Of the four establishments RAL covers the widest spectrum of the Council's involvement in research. Major facilities at RAL include a large Spallation Neutron Source, which was opened by the Prime Minister in 1986, and an internationally acclaimed laser facility.
Activities at RAL are broken down into 7 Departments. Some of these Departments, such as Administration, provide support for all RAL's activities. However, most Departments are responsible for a specific part of SERC's programme. For example, the Nuclear Physics Board's activities at RAL are supported by the Particle Physics Department while the Astronomy and Planetary Science Board's activities are supported by the Space Science Department.
Support for the Engineering Board's programme is provided by two of the seven Departments. Technology Department, as its name implies, tends to concentrate on the hardware side of the activities while Informatics Department is more oriented towards the software activities.
The Central Computing Department provides SERC's central computing service currently based on an IBM 3090-200E and a CRAY XMP-48.
The basic provision of computing facilities in universities is funded from two sources, the Computer Board and the University Grants Committee, both of which come under the Department of Education and Science. Polytechnic basic computing facilities are funded by Local Government Councils. In contrast SERC funds the additional state of the art computing facilities needed by the research community it supports.
The Engineering Board set up a Working Party in 1973 to establish the computing needs of engineers in the universities and polytechnics and to see if a rational framework might be constituted for the provision of facilities to meet these needs.
The essential findings of the Working Group were that there was a pervasive requirement for interactive computing in engineering research, that the importance of the requirement was inadequately recognised in the facilities available to universities and polytechnics and that the provision of facilities on a piece-meal basis through individual grants was not cost-effective. The Working Group recommended inter alia that the SERC should set up a centre, based at an SERC establishment, to provide facilities for interactive computing to engineers in universities and polytechnics. The centre would moreover provide new stimulus for the promotion and development of research activities over a wide range of engineering disciplines where interactive computing has a powerful role to play.
It was recommended that a small Technical Group should be set up to draw up the detailed proposals for such a centre.
The Group reported in 1975 and concluded that there was a substantial and immediate need for additional interactive computing for engineers engaged on SERC funded research. The potential user population was seen as 900 in 1975, rising to 1300 in 1980. The Report described a plan by which SERC could provide an integrated and co-ordinated service for interactive computing (the Interactive Computing Facility - ICF).
A summary of the main recommendations is as follows:
To meet the computing needs for 1976-78:
It was foreseen that by 1978, the most cost effective means of providing interactive computing would be through multi-user mini computers interconnected by a network. Multi-user minis were new (Dec had not yet announced the VAX range of machines). Existing facilities available to the academic research community offering minimal interactive computing capability were small, single-user PDP11s.
It was recommended that a review be made in 1978 to establish definite plans for the period after this.
Two multi-user minis, a Prime 400 and a GEG 4070, were purchased in late 1976 for the planned assessment following a rigorous tender exercise, which included an interactive benchmark involving teams of 6 real people running a script. This turned out to be a very revealing way of assessing the systems. An attempt was also made to make a subjective view of the machines to judge the user interfaces. In fact a P300 was originally ordered from Prime but this was changed to a P400 - the first sign that Prime regarded SERC as a very important new customer.
The Prime 400 was installed on 3 December 1976 and an internal user service started at the beginning of January 1977. The configuration of the Prime 400, which cost £81,302 including VAT, was as follows:
Prime had already decided that their machines would be limited to 8 memory slots, arguing that chip capacity would increase fast enough for this not to be a problem. In SERC's case this proved to be a very perceptive view of the future and has never caused any problems.
The Operating system at the time is best described by the following paragraphs taken from a progress report dated 1 September 1977 on the assessment activity:
The version of the operating system (PRIMOS) that was delivered was known as REV (revision) 11. REV 13 is now available and, although functionally a considerable improvement, it gives a greater store and processor overhead. For this reason we have delayed using it in service until PRIME lend us an extra 32K words of store. PRIME claim that REV 14 and REV 15 will restore the lost performance.
Because of the very rapid rate of software development within PRIME, great efforts are being taken to ensure that system modification at RAL will not lead to an unacceptable maintenance load in future revisions of the system.
The main departure from the original plan was the abandonment of the plan for a large central computer at the Rutherford Appleton Laboratory, an increase to 7 in the number of existing minicomputers enhanced and some acceleration of the Multi-User Mini (MUM) programme.
The Prime 400 was substantially upgraded when the assessment was complete and used to run a central service at RAL, supporting a user population of 100, to partially substitute for the large central machine.
The assessment of the Prime and GEC machines took place during 1977. A benchmark was produced which included a number of existing Fortran programs from users (ie Engineers). The real engineering programs were provided by the Engineering Departments of Cambridge, Glasgow and Leeds Universities and were run on the two machines using Tektronix terminals (4006, 4010 or 4014) via 1200/75 bps modems on the public telephone network.
The recommendations from the assessment were as follows:
In a non-demanding environment, the GEC 4070 is more cost-effective than the P400. Therefore, GEC 4070' s should be purchased to go to sites with this characteristic workload.
However, one in six of the observed programs appear to fall into the category where large arrays are being continually accessed.
These programs will perform much better on the P400. (Note the P400 processor speed was twice that of the GEC 4070). The type of problem area where such programs have been observed include:
In environments where such programs are found, the GEC 4070 is not recommended. In this case, the P400 is much more cost-effective.
The P400 was expected to be capable of supporting up to 8-10 simultaneous users. Some specific comments from the assessment of the Prime 400 are given in Appendix 1.
The close involvement of the two manufacturers in the assessment was seen as being of considerable benefit to the whole exercise and assured the successful outcome in both cases.
A specific policy decision was made following the assessment to both distribute and support the software (operating system and applications) centrally from RAL. This was a very important and central aspect of the ICF, ensuring that engineering researchers were able to concentrate on their research. A machine from each supplier was to be provided at RAL in support of the systems development activities.
The acceptance of Prime as one of the recommended suppliers resulted in one of the existing mini-computers to be upgraded being a Prime 300 at Nottingham University. This was replaced by a Prime 400 with the Prime 300 processor being transferred to the Rutherford Laboratory as a possible development machine. In fact it was put in a cupboard and never used!
All the other hardware recommendations were implemented according to the Technical Group Report, with one additional item, namely the purchase of a Floating Point Systems AP-120B array processor for assessment. This was attached to the P400 at RAL.
On the applications side, Special Interest Groups were established covering the following areas:
The Groups were charged with identifying essential applications to be purchased or developed during Phase 2 of the ICF.
During the assessment stage a number of 'essential' changes and additions were made to PRIMOS. These included:
A list of future developments was also identified which would be implemented in the second phase of the Interactive Computing Facility (1979 to 1984). These were:
Some of the remaining deficiencies, for example batch, X25, incremental dumping and filestore budgeting, were being addressed by PRIME in 1978.
The review of Phase 1 of the Interactive Computing Facility (ICF) concluded that the concept of providing a distributed interactive facility for use by engineers as defined in the Technical Group Report was proving successful.
Consequently a second phase covering the period 1979 to 1984 was proposed and approved. The overall objectives were:
To achieve these the Interactive Computing Facility had to:
The details of the plans for Phase 2 were widely distributed to the Academic Engineering Community and, in particular, Institutions were invited to submit bids for the provision of a multi-user machine, either a Prime or a GEC.
Successful applicants were provided with the appropriate machine and a six-year support contract. The contract covered the full cost of maintenance, the salary of a systems manager, travel and consumables for three years and then reduced linearly to cover maintenance only by the end of the fifth year.
Despite the additional case required to justify the provision of a Prime machine, five successful bids were made in 1979 and the first remote Prime, a P550/1, was installed at Sussex University in October 1979. Four further P550/1 machines were installed at City University, Surrey University, Warwick University and University College, London in 1980. Additionally an existing P300 machine at the University of East Anglia was upgraded to a P400 in 1979 and a second Prime service machine, a P400, was installed at RAL in June 1979 to serve as a development machine and provide a service to local users.
Twelve GEC 4000 Series machines were installed in the period 1979 to 1981. There was also a review of the DEC10 KI services at Edinburgh and UMIST in early 1980 which resulted in the machine at Edinburgh being upgraded to a DEC10 KL and the UMIST machine being replaced by a P750. UMIST also took on a responsibility, under contract to SERC, to provide user support for the Prime community.
The remainder of the Phase 2 period saw a steady consolidation of the Prime hardware and the associated systems and applications software. A planned programme of memory, disc and processor upgrades reflected both the increasing number of users of the Prime Service (1000 by 1981) and their increasing demands on the computing resources. Further development of PRIMOS concentrated on the provision of X25 networking to ensure that all the Prime machines were connected to the SERC X25 Network.
The size of the terminal pool had been increased to 1000 (50% being graphics terminals) by 1982 and these were widely distributed to grant holders throughout the UK. In many cases terminals were connected to remote machines via time-division multiplexed leased lines.
By the end of 1981 there were ten Prime systems in the ICF, 3 × P750, 5 × P550 and 2 × P400. The expected number of simultaneous users which could be supported were:
P400/P550 [1 Mbyte memory] 10-15 P750 [1. 5 Mbyte] 15-20
The good experiences of the ICF with Prime machines resulted in a number of other projects purchasing Primes (details of these are given later in this paper).
All these systems were connected to the SERC X25 Network and offered both interactive terminal and file transfer facilities. Support and distribution of the PRIMOS operating system was carried out via the network. All Primes were able to use the file transfer facility to access the large central IBM facility at RAL for computationally intensive activities.
Figure 1 shows the distribution of the ICF computers in the UK in 1981.
The AP-120B was transferred to the P750 at RAL when it replaced the original P400. Although a few users made effective use of this array processor the overall assessment concluded that there was great difficulty in using the machine because of inadequate software and the AP-120B was taken out of service in 1983.
The applications activities on Prime during Phase 2 concentrated on Control, Electromagnetics, Finite Elements, Digital and Analogue Circuit Design, Pre- and Post- Processors, Graphics and Database Management Systems. It had become clear by 1981 that the Primes were far more effective than the GEC machines for running real applications (thereby confirming the recommendations from the assessment programme in 1977). GEC users requiring to use the Control, Electromagnetics, Finite Elements and Circuit Design packages had to access a Prime over the Network.
The SERC version of PRIMOS was stabilised during this Phase of the ICF. A major problem of making such major changes to the standard PRIMOS was that it took at least 6 months to incorporate them when a new Revision was received. In practice this meant that the ICF did not take every Point Release of PRIMOS. Several unsuccessful attempts were made to get Prime to accept the SERC modifications during this period and thereby provide support for the ICF operating system on the Prime machines. Needless to say there were a large number of demonstration disputes with Prime over the years!!
New software additions to all the Primes in 1984 included PROLOG, EMACS and the C Compiler.
A new element appeared in the computing facilities offered by the Engineering Board to its users in 1982, namely the planned introduction of Single User Systems (ie workstations) as part of a Common Base Policy. By the end of 1984, these systems were beginning to reduce the demand for ICF resources. Strategically, from SERC's point of view, Single User Systems were now the state of the art computing systems and Multi-User Minis were now run of the mill.
A further problem in 1984 was that there were severe financial pressures on SERC which were expected to remain for several years.
The immediate effects of the financial pressures on SERC was to force the ICF to review the number of machines it supported and to cease support for those which were either not cost effective or, in the case of the DEC10 at Edinburgh, were high cost (the DEC10 cost six times more to run than the average GEC or Prime machine). The GEC Operating System, also heavily modified by the ICF, was frozen in 1985.
In 1985 the Prime at Nottingham, the DEC10 at Edinburgh and six GEC machines had their ICF support withdrawn. Nottingham University decided to continue to run the Prime at its own expense and upgraded the machine to a 9655 later in 1985. ICF continued to provide and support PRIMOS for this machine via the Network.
Continuing financial pressure coupled with the success of the Single User System Programme with the appearance of Sun Microsystems resulted in further machine closures during this period. By the end of 1987 there were in excess of 400 SUN workstations supported by the Engineering Board throughout the UK. The Computer Board also began to fund interactive computing facilities during this period and in many cases these were able to provide the necessary resources for SERC grant holders.
Inevitably demand for the still large ICF Prime resources began to drop and between March 1986 and May 1988 all remote machines had their ICF support withdrawn. The machines at UMIST and Warwick were taken over by the University and will continue to run for at least two years. The only ICF Prime now running is at RAL. This was upgraded to a P9955 from 2 × P750 in 1986 and will run until March 1990. Support services have now been reduced to a minimum level. The system currently has 200 users. The SERC version of PRIMOS was frozen at Rev 20.2.4 in 1987.
The GINO-F Graphics system was frozen at level 2.6 in 1985 and the Graphical Kernel System, GKS was introduced. By September 1987 full ISO GKS Level 2b was available on the remaining ICF Primes. Applications developments were also frozen in 1986 although the central support for the existing packages was still available. Appendix 2 lists the Engineering Applications packages available on or from the ICF Primes in 1986.
Table 1 gives a total summary of the ICF Prime machines for the period 1976-1988.
| Location | Initial Machine | Intermediate Machine(s) | Final Machine | ||
|---|---|---|---|---|---|
| Date | Type | Date | Type | ||
| RAL (Services) | 1977 | P400 | 2 × P400; 2 × P750 | 1986 | P9955 |
| RAL (Dev) | 1977 | P300 | P400; P550/1 | 1986 | P9655 |
| Sussex | 1979 | P550/1 | - | - | - |
| City | 1980 | P550/1 | - | 1986 | P750 |
| Surrey | 1980 | P550/1 | - | 1986 | P550/2 |
| UCL | 1980 | P550/1 | - | 1986 | P750 |
| UMIST | 1980 | P750 | P9950 | 1985 | P9955/1 |
| Warwick | 1980 | P550/1 | P550/2 | 1986 | P750 |
| East Anglia | 1979 | P400 | - | 1986 | P750 |
| Nottingham | 1977 | P400 | - | 1985 | P9655 |
Table 2 summarises the typical final configurations by machine type.
| P550 | P750 | P9955 | |
|---|---|---|---|
| Memory (Mbytes) | 1.0 | 2.0 - 4.0 | 8.0 - 12.0 |
| Disc Space (Mbytes) | 500 | 1200 | 2700 |
| Magnetic Tape Unit (9 track) |
1 | 1 | 1 |
| Printers | 1 | 1 | 1 |
| Plotters | 1 | 1 | 1 |
Many university research groups collaborate closely with industry, particularly in the areas supported by the Engineering Board. The following three activities are given as excellent examples of successful collaborations which have made heavy and effective use of the ICF Primes and in two of the three cases the companies involved purchased their own Prime machines:
The first commercial package developed on the ICF Prime was SAMMIE (a System for Aiding the Man-Machine Interaction Evaluation). The development started on the P300 at Nottingham before the ICF started and the commercial package, completed on the ICF P400, was marketed worldwide by Compeda Limited. Prime, of course, took over Compeda Limited in 1982 and now share the worldwide marketing rights for the Prime version with SAMMIE-CAD.
The success of the ICF Primes and the special centralised support arrangements provided led to a number of other Prime facilities being purchased from other funding sources. By special arrangement all these machines were integrated into the ICF network and provided with the same support services as the standard ICF Primes. By 1986 all of the service machines at RAL were connected via Ringnet and shared common archiving and back-up facilities.
The following activities purchased Primes:
This initiative was specially funded by the Engineering Board and introduced a number of Primes between 1980 and 1986. The Initiative established a number of MSc courses and funded grants in universities and polytechnics on the design and prototype manufacture of Integrated Circuits. The Primes were provided to support these courses and ran the GAELIC package. The actual Prime facilities were:
In 1982 RAL decided to evaluate Prime's Office Automation System (OAS) and purchased a P550 for this. In the event RAL decided to implement IBM PROFS for its Office Automation. The P550 was transferred to the Library as a service machine running INFO and STATUS. This proved to be a successful service and an upgrade to a P750 was effected in 1985.
In 1981, with generous support from Prime (UK), a P400 was installed to provide a service for five local Secondary Schools. Terminal facilities were installed in the schools and connected via time-division multiplexed leased lines to the Prime. This has been a very successful project and the P400 was upgraded to a 550/1 in 1986, again with generous support from Prime. One of the benefits of this project to RAL has been a significant number of recruits, straight from school, with experience of using Primes.
In 1982 RAL purchased a P750 as the focus for its introduction of CAD into the Laboratory. This ran both MEDUSA (from CIS!) and DRAGON (a wire-wrap package). By 1984 CAD had proved so successful in the Laboratory that a second P750 was purchased to enable MEDUSA and DRAGON to run on separate machines. A further upgrade to replace the two P750s by a P9955/2 took place in 1986. RAL is currently discussing the future of this CAD facility with Prime and is interested in a Prime-Sun solution.
A total of four P2250s have been purchased directly on Engineering Board grants, two at Bath University, one at Imperial College, London and one at RAL. The RAL machine was upgraded to a P2655 in 1986 and to a P2755 in 1987. This machine is currently used to support a research grant funded through the European Special Programme for Research in Information Technology (Esprit).
The provision of Prime computers as part of the Interactive Computing Facility has proven to be an unqualified success for the SERC's Engineering Board. In the twelve year period between 1976 and 1988 a total of approximately 10000 users have benefited from the service offered on the Prime machines. The estimated total cost of the ICF Prime service, including capital, maintenance and manpower, over this period is £12M. The ICF allowed researchers to concentrate on their research and enabled innovative research projects to be undertaken which would not otherwise have been possible.
As an added bonus, the success of the ICF led to the purchase of a further highly effective 12 Prime machines by SERC including the establishment of the successful Schools Project based on machines donated by Prime.
The only significant problem with the initial configuration was the 80Mbyte discs which gave problems at a rate of about one a week. Eventually PRIME replaced them both and the new ones have given no trouble. The only other problems have been with upgrades whether major (for example new memory) or minor (CPU modification). Invariably there is a period of unre1iability (one of two incidents a week) with such upgrades. After this there is no further trouble, and it is believed that a stable system would give no more than one incident a month. However, there have been a number of lineprinter paper jams. Software breaks on the PRIME are virtually unknown.
These figures represent the number of failures which have caused the systems to halt during service periods over the 12 months of the assessment.
PRIME GEC
Hardware: memory 19 12
disc 26 3
other 25 0
Software: 10 25
Most of these incidents were clustered in time, and it is common to have several weeks with no incidents on either machine.
Generally these machines are very reliable compared with any mainframes previously experienced.
The standard of documentation on the PRIME used to be very poor: only in the last few months has good quality documentation, orientated towards the user, begun to emerge. It was necessary to produce a substantial user manual for the machine.
PRIME automatically supply the source of all software: this makes it easy to enhance, or to cure trivial bugs in utilities, locally.
Initially PRIME software support in the UK suffered because of the remoteness from the parent company. This appears to be no longer the case: in fact a UK development organisation is being set up. Relations with Prime have been excellent, although RAL appears to be treated as something of a special customer.
Engineering Applications Packages available on or from the ICF Primes (1986).