In considering the future organisation of SERC computing facilities we have reviewed their development and reviewed SERC's role in the support of computing for university research in relation to the role of the Computer Board and the UGC. The following section gives a brief account of the support given by each of these funding bodies and gives some information on the historical development of SERC computing facilities.
Within the UK there are three different funding routes for the support of computing for university research. These are: through the university using UGC funds or funds directly available to the university, department or group; through the Computer Board either by support for facilities in the university or through the Computer Board's National Centres; and lastly through the Research Councils either by use of their central facilities or by the provision of facilities directly in the universities. Not all Research Councils have the same policy for the direct support of computing for university workers and SERC in practice is by far the largest provider of such resources among Research Councils.
It is difficult to provide precise information on the level of support provided through these three routes but the following generalization will give the correct overall impression. The university route is by far the largest provider but since each university, and in many cases even each department or group, has its own independent control over its computing prevision, this route is completely uncoordinated. If the full costs, including the cost of staff providing the computing service, are included the scale of the provision is about £50-100M per year. The number of staff involved is large - probably well in excess of l000.
The second largest provider of resources is the Computer Board with a financial provision of about £30M per year. The secretariat staff available to the Computer Board is very small (about 6) which limits the degree of direct technical support and of coordination.
The SERC currently spends about £16M per year on the provision of computing services for university research with somewhat more than half providing facilities supported through the SERC establishments but largely distributed in the universities and the rest being used for direct funding of systems in the universities largely through the grant mechanism.
Computing at the Chilton site started in the early 1960s with the installation of Ferranti systems at both the Rutherford High Energy Laboratory (RHEL) and the Atlas Computer Laboratory (ACL). RHEL installed an Orion which was used for Nuclear Physics Board work while ACL installed the largest UK machine, an ATLAS. This provided facilities for the Harwell Laboratory and a national-centre-like facility for universities. It also provided support for RHEL and for those activities which now come under the Science Board.
The need for more data processing capacity by the High Energy Physics community resulted in the Orion being replaced by an IBM 360/75 in 1967 and a 360/195 in 1971. An ICL 1906A was installed in the Atlas Laboratory at the end of 1971, and the Atlas computer was closed down in 1973.
In 1975, ACL and RHEL were merged into a single Laboratory (RAL) with the bulk of the Science Board computing work being moved either to the Computer Board's National Centres or to Daresbury. A second 360/195 was added and the ICL 1906A phased out in 1978.
Since 1978, RAL has been going through a replacement programme which is still not complete. This was started by adding a front-end processor, an IBM 3032, to aid the conversion. The main changeover then took place in 1982 when both 360/195 processors were replaced by an IBM 3081D and an ICL ATLAS 10, phasing out the 3032 in the process. The main peripheral equipment has still to be replaced.
From 1977, RAL was also active in setting up the Interactive Computing Facility (ICF) which resulted in the Engineering Board funding a large number of multi-user-mini systems which were sited in university departments for interactive engineering applications. More recently, the Engineering Board has been in the forefront of a move to the use of Single User Systems (primarily the ICL PERQ) as professional scientific workstations.
The first mainframe to be installed at Daresbury was an IBM 360/50 in 1966. This was replaced by a 360/65 in 1968, a 370/165 in 1973 and an IBM compatible AS7000 in 1981. In the early years the machines were used mainly for high energy physics work associated with experiments at Daresbury and CERN. This was phased out from 1976 to be replaced by Science Board computing and increasingly by work related to the Nuclear Structure Facility (NSF) and the Synchrotron Radiation Source (SRS).
During the 1970s there was growing pressure for the provision of a very powerful number cruncher for large scale calculations in theoretical science and for modelling. The needs were mainly in areas covered by the Science Board (atomic and molecular physics, condensed matter, quantum chemistry, etc) and were met in 1979 by the prevision of a CRAY 1 computer at Daresbury. In 1983 the CRAY was moved to ULCC to provide a national facility funded by the Computer Board.
SERC has played a central role in the development of networking for UK universities. The activity started in the early 1970s with the formation of star networks of remote job entry work stations from RHEL and the Atlas Laboratory. Similar networks grew with the formation of the ICF and towards the end of the 1970s these separate networks were drawn together into a single entity, SERCNET, through the development and implementation of standard protocols. This work was jointly funded with the Computer Board and has resulted in agreement to support a unified network for the whole UK academic area - JANET.
The total number of university users of SERC centrally supported systems is about 5500. It is not possible to quantify the number of users of systems installed through the grant mechanism but it is believed to be comparable.
Appendix A gives statistical information on the computing provision made by SERC.
The Computer Board was set up in 1966 to carry forward on the basis of planned development the proposals for providing computers for research in universities and Research Councils which were set out in the report of the Joint Working Group on Computers for Research (Cmnd. 2883, January 1966). The terms of reference of the Board specify that it should, on the basis of a continuing review of needs, make recommendations to the Secretary of State for Education and Science in respect of the provision of computers to universities. The Board is also required to advise Research Councils on their computer proposals (including significant peripheral equipment, but excluding computers provided solely and essentially for the purpose of specific research projects). These functions remain central to the work of the Board.
A large part of the Board's programme is concerned with the support of two National Centres - the University of London Computer Centre (ULCC) and the University of Manchester Regional Computer Centre (UMRCC). The National Centres provide the following services:
Both National Centres have recently taken delivery of supercomputers provided by the Board which will help to keep the facilities at these centres at the leading edge of computer technology. In determining the scale and style of the facilities at these centres, the Board has worked closely with the SERC to coordinate the provision of systems to meet the needs for specific research and for more generalised usag0 from within the university community.
Another national facility provide] by the Computer Board as an aid to research is the Distributed Array Processor at Queen Mary College, London. This has proved a valuable facility for handling many types of problem.
The Board is supporting a network programme to rationalise and enhance existing university and Research Council computer links to provide a modern packet switched network which will offer a unified wide-area data communications service for the academic community (JANET).
The Computer Board provides individual university institutions with mainframe replacements on a ten-year cycle, with intervening enhancements at the mid-term point. It is up to the individual university to set out its research needs within its operational requirement for the replacement or enhancement of its system. Insofar as it is possible and practicable, the Computer Board endeavours to meet those requirements in terms of an appropriate level of grant. However, where very specialised and costly areas of research are involved, such as for example X-ray astronomy, the Computer Board hopes that support from other bodies, such as the Research Councils, will be forthcoming. Additionally, the Computer Board does not fund equipment for departmental research as that is a matter for UGC support. It does consider funding central computing facilities designed to supplement those provided in a university department if it can be demonstrated that the enhanced configuration will then offer a service to the university as a whole.
Due to the nature of Computer Board funding and the mechanism for the disbursement of its funds, it is not possible to identify precisely the support which is committed specifically to research. Demand on funds has been, and remains, heavy. For the financial year 1983/4 the Computer Board is operating within cash limits of £14.5M for capital grant and £18.0M for recurrent grant. For 1984/5 the cash limits will be £12.7M and £18.9M respectively.
The University Grants Committee, after advice from its Equipment Sub-Committee, provides universities with a grant for equipment and furniture from which departmental computing equipment and administrative or library computing equipment (including word processors) may be purchased, subject to certain approvals. The costs of providing data transmission facilities via digital telephone systems may also be met from the equipment grant, subject to the approval of the Committee. All other recurrent costs of central and departmental computing services at universities, including the maintenance of equipment purchased from UGC funds (but excluding costs which are met by the Computer Board) are met through UGC recurrent grant. The latter may also be used to fund the replacement of telephone switchboards and links.
The past decade has seen a continual increase in computing power per processor, with a performance/price ratio growing at a rate of between 1.2 and 1.5 per year. (Figure 4, Appendix A) This rate of growth is even higher for small systems. The increased processor speeds at the lower price end of the scale will accentuate a trend towards more distributed systems. However, there will continue to be a need for large central mainframe systems because of the high data- processing load generated by SERC's various research fields, and by the need for university researchers on a number of geographically separated sites to access the same central database.
Those researchers whose primary tool for theoretical calculations is a computer will continue to demand the fastest systems which can be afforded. The rate of growth in processor power is greater at this level than at any other part of the range. The next decade will see a greater variety of machine architectures becoming available ranging from the vector enhancements to IBM-compatible processors through vector processors such as CRAY and CYBER to multi-processor configurations and more novel architectures.
Multi-user-mini systems are now the most common departmental systems in universities. It is less clear whether this mode of working will continue to be the primary method for a researcher to get his interactive facilities. It is even less clear whether SERC should in future be involved with such provision. The future of the current multi-user-systems is more likely to be as filestore engines or as specialist processors in a fully distributed environment, where the main interactive facility for the researcher will be a single user professional workstation available in his office or place of work.
The growth in this area of the market is by far the greatest. From only a handful of companies offering products two years ago, there are probably over 200 vendors today. Of particular importance to SERC is the maintenance of its Common Base Policy in order to ensure a good software environment for such systems. At the lower end of this market, the software costs often exceed the hardware costs.
Of particular significance here is the appearance of very powerful personal computers such as the IBM PC/370. The latter can emulate many of the front-end systems which are currently used by the researcher on the mainframes. Consequently it will be eventually possible to devolve a great deal of the central computing workload to the user terminal, thus freeing the expensive mainframes for the bulk data processing load which is their major function.
The establishment of JANET should provide a much more integrated computing network for the university community. However, there will be a great deal of work to be done in the next few years particularly in harmonisation of the existing networks, in increasing the bandwidth, and in moving to international standards as they appear.
The way in which Local Area Networks will develop is much less clear as international standards for hardware and protocols are only just beginning to appear. Significant changes are likely to appear in these over the next few years. In particular, the standardisation of the token ring and announcements from IBM are both likely to be important.
A major development of great importance to SERC will be the appearance of optical discs which will allow much larger databases to remain on line for long periods. This is of particular significance to SERC where much of the experimental data is read-only. Current on-line data storage devices cannot approach the capacity needed in a typical data processing run from, say, an HEP experiment.
In software development the move to standard operating systems is beginning. The impetus for this initially came from the personal computer market. The appearance of UNIX on a wide range of systems including large mainframes, vector processors, and single user systems will have a major effect in increasing the portability of applications and personnel between one system and another.
