Prior to reaching any firm conclusions as to the solutions to the problems outlined above, the Working Party took evidence from a number of bodies, including the Boards, as to future needs for SERC computing, particularly over the next few years. In holding meetings with users, and in inviting submissions, discussion was not restricted to the future usage of existing machines but comment was invited on all aspects of computing, including the administrative arrangements for its funding within SERC and the provision for research computing by the Computer Board and the UGC. The summary of discussions held and of submissions considered is outlined below. We do not necessarily agree with all of the views expressed in the following sections but have taken note of them in drawing up our recommendations. The various submissions received are given in Appendix B.
The ASR Board envisages a trend away from the use of central facilities towards an increased use of distributed computing, particularly in work in astronomy which will increasingly be undertaken using the Starlink system. There is a consequent requirement for an effective network, not only for distributed computing but for the remote operation of telescopes. Workers in theoretical astrophysics will, however, continue to require access to large-scale processing, and in this respect the Board would like access to a large state-of-the-art computer with appropriate peer review. The ASR Board has also noted that there is likely to be a large requirement for computing in geophysics with a use of both distributed computing and of central data storage on an appropriate computer. A continuing, but small, amount of time on the ICF for engineering design work is also required for the support of space and ground based projects.
The Engineering Board identifies an evolving requirement for powerful personal workstations and for single-user systems backed up by appropriate networking arrangements, and considers that, as the ICF no longer always provides the right kind of facilities for its users, there should be a change in the way that this particular project is run in the future. In this context the Engineering Board considers that it would be appropriate for it, as the major user, to take over the management of the ICF and the single user systems programme. The Engineering Board sees the support for, and provision of, applications software and software packages as of the utmost importance, and it would wish to see a continuation of central control of this area, with an increased provision in the future. The Engineering Board also identifies a convincing need for central batch facilities and a requirement for an effective network.
The requirements for powerful central mainframe computing for particle physicists are highlighted in the submission from the Nuclear Physics Board, who also propose that the responsibility for funding the mainframe service might be transferred to the NP Board. The NP Board would then provide a commercial service for the 25% use made of the mainframe service by other SERC Boards. The NP Board envisages that appropriate utilisation will be made of the developments in ultra-fast array processors and distributed interactive computing as possible alternatives to vector processors such as the CRAY, and in this respect the Board highlights the importance of networking. It would also wish for common European standards to be implemented for hardware and software. For nuclear structure physics, the NP Board envisages that data processing could, in future, be carried out by dedicated computers at Daresbury and at universities, with a consequent decline in the need for central computing facilities. A small amount of central batch processing would, however, still be required for the theory groups in this area.
The Science Board envisages an increased requirement for the collection and analysis of data from its central facilities, and it recommends that the AS 7000 computer at DL should become wholly dedicated to SRS data processing with Science Board assuming responsibility for it. The Science Board also identifies a continuing need for access to a large state-of-the-art computer. In the short term, the Science Board requires good access to the CRAY and CYBER at the National Centres, and it would wish for this to be on a peer-review basis. In the longer term, the Science Board hopes that SERC will provide the next generation of advanced computer, which would be used almost entirely for peer reviewed computing. with regard to grant-related computing, the Science Board has identified a continuing requirement for large-scale batch processing and for associated central support which cannot, at present, be met from CB funded facilities. The Science Board also considers that the central co-ordination of software will become increasingly important.
The Board submissions are in Appendix B.
The users of the Daresbury computing centre, who included past users of the CRAY at DL, considered that SERC had a role to play at the forefront of technological developments in computing and should not necessarily provide only a routine computing service. These users felt that they must have access to vector processors, and they felt strongly that SERC should, in the short term, continue to grant-aid large scale use of vector processors and, in the long term, consider installing the next generation of super-computer. All users considered that an effective network was very important, and that its development should be fully funded. The networking environment was seen to be a crucial factor in the success of the collaborative computational projects.
The nuclear structure physics users at Daresbury considered that there was a real need for a computing service at Daresbury, primarily to provide expertise in computing problems. The requirements for data analysis could better be met by a small number of dedicated computers on site linked to university sites by an effective network. This would allow large-scale interactive processing of data which at present was not possible on the AS 7000, without causing disruption to the other services that the computer provided. The small amount of computing associated with theoretical work could be carried out at either site.
The SRS Working Party report highlighted the trends it saw as likely in this area. The data-processing requirements from the SRS were likely to expand rapidly and at least one quarter of the data generated needed to be processed on site, which would occupy most, if not all, of the AS 7000 and would require some additional VAX-like computers. The SRS Working Party had no confidence that British Telecom megastream links would allow work to be carried out at RAL, as they were yet unproven.
The Joint User Liaison Committee presented their views to the Working Party and submitted a paper which is at Appendix B. This Committee believed that it was imperative that computing was science driven and that fashionable hardware was not bought without there being a valid scientific case. There would be a continuing need for central computing facilities whose planning should not be jeopardised by an unstable funding situation. There was a strong feeling in the scientific community that the SERC should provide the next generation of super-computers and that the provision of a first-class computing capability should be of first priority rather than necessarily the country of origin of the computing equipment. With regard to computing support, users believed that this should be interest-based rather than geographically centred. Users believed that the possibility of devolving responsibility for facility-related computing back to Boards should not be overlooked, and some users also supported the idea of the Nuclear Physics Board taking over responsibility for the RAL batch service.
The RAL users who were present at the meeting with the Review Working party were generally satisfied with the computing support provided at RAL. All these users wished to see some form of stability established in the computing service, with good forward planning and adequate consultation with the user community. Examples of specific difficulties were presented: the SNS users were concerned that there would be an inadequate provision for data analysis when the facility was fully operational; DEC and GEC users said that they would have wished for more consultation and adequate planning before withdrawal of support for GEC and DEC machines took place; and ASR Board users present at the meeting said that the provision for their area was inadequate following the ASR Board cuts.
For the future, the HEP users wished to see an adequate provision for large-scale computing which should be done centrally, but they proposed that responsibility for this should be given back to Boards. There was general support for the idea that facility-related computing should be the responsibility of the Boards; ASR Board users believed that a mechanism should be found whereby project funds could be diverted to computing if this was felt to be necessary. If such ideas were accepted, users believed that computing provision would then be science-driven, but they suggested that Boards would need to take more account of users' views. Users also believed that there should be direct Council support for general infrastructure items, which would include networking. The particular items included in the infrastructure should be subject to control by users. The need for increasing co-ordination of software development, and for support and enhanced funding for this area was emphasised. A central software service should include the co-ordination and provision of state-of-the-art software and software techniques, and provision for training in the use of software techniques together with a consultancy service for software problems. Users were conscious that Boards might overlook the high costs of providing such software support. Likewise, networking was also considered to be of high priority, but it was considered that it might not receive adequate funding if funded directly by the Boards.
Some users strongly supported the case for SERC providing the next generation of supercomputer, but others believed that an alternative would be the development of array processors which could be attached to local computers.
The majority of SERC users of the CRAY service at ULCC expressed dissatisfaction with the present access arrangements and, although recognising that these would improve, considered that the Computer Board approach to the provision of vector processing would never, in their view, be satisfactory for large scale research use. In the view of such users, it was felt to be a Research Council responsibility to provide for specialised state-of-the-art computing which should be accessed on a peer review basis. Provision for the next generation of large scale state-of-the-art computing should be considered in the immediate future, and SERC should ensure adequate planning not only for this but for other equipment purchases. The mechanism for funding such computing provision should be such that it was not disrupted by short term financial crises. SERC should also look carefully at the impact of technological developments and consider carefully whether it might be cost effective to provide more local computing facilities; this would enable the large scale computers, whether SERC or CB funded, to be relieved of small and inappropriate jobs. Users also supported an increased central responsibility for the co-ordination and standardisation of software, and highlighted the increasing importance of wide and local area networks.
Whilst the views of the community involved in collaborative computational projects (CCPs) were not directly sought by the Review working Party, it received a number of letters from Chairmen of these projects.
The Chairmen of these projects believed that the co-ordination of the development of software as achieved by the CCPs was very important. Several mentioned the high value that they placed on the support from staff at DL and RAL, and the importance of the development of networks. It was generally believed that such central support should continue, whether funded directly by Boards or through the CCC. Several Chairmen mentioned the heavy reliance on good access to large-scale central processing particularly the CRAY, and hoped that access to the CRAY and CYBER should continue to be funded through the SERC. Some of the CCPs also strongly supported the idea that SERC should provide the next generation of large-scale state-of-the-art computer, or at least guarantee the peer-reviewed access for large users to such a machine. Several also expressed the view that the Computer Board could not provide such specialist facilities for research purposes.
The Steering Committee on Administrative Computing and Office Automation presented a paper to the working Party which reviewed the progress made in developing proposals for a computing system to satisfy the increased requirements of administrative computing. This paper is included in Appendix B.
The growth in non-scientific computing and the introduction of modern office technology and electronic office concepts has long been identified, and there is already extensive use of computing for administration, but there are not yet any common or compatible systems implemented on a Council wide basis. The Steering Committee considers that the introduction of modern computer based administration systems will increase the efficiency and effectiveness of the work of the Council and its Establishments and would lead to greater productivity of staff at all levels, better control over administrative processes and some saving in clerical effort.
A number of Council-wide initiatives have been launched and the Steering Committee has identified a requirement for computing systems underpinning these initiatives to be VDU-based, interactive, flexible, and user friendly. The Committee are currently considering two options for provision of such a system: the use of a separate new machine with a capacity of around 2-3 MIPS, and the use of the present scientific machines. The latter option would involve either placing all the administrative front end activity on the 3081 which might involve substantial operating systems conflict or loosely coupling the 3081 and Atlas 10, providing a front end processor for the Atlas 10, and using the 3081 exclusively for administrative computing during the day.
A number of problems have been identified which would have to be overcome in providing a computing system for administrative computing. Any system would need to be secure, reliable, provide sufficient resource for a large number of simultaneous interactive users and be capable of enhancement when the predicted increase in administrative work load occurred. The Committee prefers the idea of a separate administrative machine, but recognises that this might carry a cost premium. If the scientific machines were to be used the Committee would prefer the loosely coupled option to the alternative of a shared two machine complex with a common front end.
Over the next several years, computing facilities for research will be increasingly provided by a range of powerful, relatively cheap, single-user computers offering a greatly improved user interface. On any campus, these machines will be linked by fast local area networks, giving access to shared facilities such as bulk filespace and a high quality printing service. Due to the wide range of equipment available, it is important to focus, as far as possible, scarce research-support activity onto a limited range of equipment. This range should be centrally supported, with funding much above the currently inadequate levels, to enable the provision of a rich range of software providing both program development tools and a wide variety of applications software.
