Dr Howlett said that he was sure that the Committee would wish to congratulate Mr D W Davies on winning the John Player computer award for 1974. The Committee endorsed this.
1. The Minutes of the last meeting were confirmed. The Chairman thanked Dr Thomas for standing in as chairman at the meeting.
2. Dr Thomas asked when Council's Forward Look included provision for a major machine enhancement. Mrs Paton said that there was provision in the Council's Forward Look for a number of capital projects. There was a considerable degree of flexibility in the timing of these projects since funds had not yet been earmarked for them. The Chairman said that the validity of the Council's Forward Look depended on the future financial constraints, the readiness of capital projects and the priority assigned by Council to them.
3. Dr Howlett reported that the contract for the FR80 microfilm recorder had been settled. There was a risk that some expenditure might slip into the next financial year because the manufacturer, III (Information International Inc), was experiencing difficulty with some subcontractors. The company had agreed to provide a definite time scale for the equipment by 15 January 1975.
4. Dr Howlett reported that the Science Board had deferred a decision on the provision of two GEC 2050 terminals and wished to have the views of the Computer Review Panel and of the IBM 370/195 Management Committee. The latter committee had met the previous day and had raised no objection to the proposal. The technical group of the Computer Review Panel had raised no objection to the proposal although the Computer Review Panel itself had not yet met. Mrs Paton said that the matter would be put to the Science Board at its meeting next week.
5. Dr Howlett reported that the telegraph scanner which had been approved at the meeting in May 1974 had been ordered from ICL. Although delivery had been quoted as August 1975, it was now hoped that ICL would deliver the equipment in March 1975.
6. Dr Howlett commented that the new building had proved a great success and the Colloquium Room, where the meeting was being held, was proving a great asset. Five of the six rooms obtained from the conversion of the old conference room had been handed over. The sixth had been delayed because of extra work needed in the removal of a fire escape.
7. Dr Howlett reported that the ICL 1906A computer was now running very smoothly. Recently ICL had undertaken a major job overhauling the connectors on the machine. This was a measure taken to prevent the onset of faults arising from metal migration found in the connectors on other ICL 1906A's and to a very much lesser extent on the machine at Chilton.
8. Dr Howlett said that the software in use on the 1906A was George 4 Mark 7.7. It was proving extremely reliable. There were plans to change over to Mark 8 of the George 4 operating system but this was not expected to cause a major upheaval since it was largely a consolidation of the existing version in a tidier, more manageable and compact form. It would allow further changes and corrections to be made more easily than in earlier marks. It was expected that this would be the final version of George 4 supported by ICL. The Laboratory wished to continue deriving benefit from the support given by ICL for its software.
9. Dr Howlett reported that the distribution of CPU time on the 1906A gave rise for some concern. The amount of idle time and the amount of time spent in George organising the workload and filestore activity seemed to be entirely acceptable. The Laboratory was concerned that Executive took rather more time than seemed reasonable. Representations had been made to ICL and there was a possibility that with George 4 Mark 8 there might be some improvement in Executive. It was felt that if 10% of the time were used by Executive this would be reasonable for the activity undertaken. It was possible that the use of time on this scale was a consequence of the machine architecture. Dr Howlett confirmed that the distribution of CPU time usage on the day shift was not the same as at other times.
10. Dr Howlett reported that the interactive graphics system was now available and working. A manual had just been prepared describing the system. A seminar would be held on 31 October to inform users of all the graphics facilities available and also of the new microfilm recorder.
11. Dr Howlett said that the Laboratory's involvement in communications was well demonstrated by the list of terminals shown in the paper. It also illustrated the change in the pattern of access to computer facilities. He said that the Laboratory hoped shortly to begin experimental use of the ARPA Network to allow Professor Walkden of Salford University to carry out fluid dynamics calculations on the Illiac 4 machine at AMES Research Centre in California. The Laboratory also would be obtaining connection to the Post Office EPSS Network and since it was expected that future development of communications networks would be along the lines of EPSS this was an important development.
12. Dr Thomas said that the terminal at the NERC Institute of Geological Science in Kensington had been omitted from the list. He also wished to register approval for the help given by the Laboratory to Glasgow University in permitting use of the 1906A by means of a link through the Royal Observatory Edinburgh. This work had been valuable in preparing for the installation of the ICL 1906S computer at Glasgow University
13. The Chairman reminded the Committee that the Ariel 5 satellite had been successfully launched on 15 October 1974 and was now in orbit. Dr Howlett said that the Laboratory would be involved in providing standby facilities for the control of the satellite.
14. Mr Fossey reported that the processing of the data from the S2-68 satellite system was being undertaken on the IBM 370/195 computer. About one third of the data had been processed and it was hoped that by using the computer at the level of 80 minutes per week the rest of the data could be processed by July 1975. Mr Fossey said that although the processing of the data from the Ariel 4 satellite and the S2-68 satellite would finish, the experimental teams would wish to undertake studies of the data obtained and this work would be additional to the time required for the initial processing.
14. Dr Howlett said that data banking was a most important growth area. He was extremely glad to have Professor P G Sutterlin from the University of Western Ontario who was spending a year from July 1974 at the Laboratory.
16. Mr Fossey introduced the paper by drawing attention to the average weekly usage on the IBM 370/195 computer. In the first three quarters of 1973/74, this had averaged 23 hrs a week. Following the installation of the extra 1 Mbyte of core storage on the 370/195, the Atlas Laboratory's 20% share had risen to 24 hours per week and in the most recent quarter of 1974/75 an average weekly usage of 32 hours per week had been obtained. He pointed out that the usage figures reflected use of bonus time on low priority work taken mainly at weekends. He instanced the usage by Professor Singer in the two quarters of the financial year 1974/75, where the usage had been 53 hours. and 43 hours respectively. The guarantee in these quarters was 13 hours. Similarly Professor Hockney, who now had two grants of an hour a week from the Physics Committee and from the APGC of the ASR Board, Dr Petravic and several chemists with grants from the Chemistry Committee had been able to take up substantial amounts of bonus time. This had been possible because some guaranteed users had not taken their time and because time was available for low priority work at weekends during the summer months.
17. Mr Fossey said that the average weekly usage of the ICL 1906A had risen from 35.7 hours in the first quarter of 1973/74 to 73.2 hours in the second quarter of 1974/75. Part of this increase arose from use of time at weekends but it reflected the greatly improved reliability and performance of the hardware and software in the period. Major users of the facilities had been Professor Burke's group in Belfast, which had taken about 205 hours in the two quarters of 1974/75, the Ariel 4 initial processing, which had accounted for 117 hours and several users who had used between 20 and 30 hours of whom Professor Murrell at the University of Sussex was an example.
18. Mr Fossey said that the most important matter was the growth in demand for facilities on the 370/195. When allocation of time on a guarantee footing started it was inevitable that a very substantial growth in the allocations would occur at the beginning. However by comparing the position on 1 September 1974 with that a year ago, a commitment of about 23 hours for six months compared with about 14 hours, demonstrated a growth. in demand of about 60% on the figures of a year ago. It was clear that the 24 hours a week available for commitment on the 370/195 were now more or less fully committed for about six months and that the demand from existing users reapplying for time and from new users applying for time for the first time could not be met from the existing allocation. He said that the provision of additional computing facilities was now urgent and an increase in the share of time on the 370/195 computer from the 20% level to 30% would only act as a temporary palliative to the demand. At the same time the requirements of the Nuclear Physics users were also increasing.
19. The Chairman said that this was impressive evidence of the way in which demand for computing was developing.
20. Mrs Paton asked if the Committees had been made aware of the way in which time had been used by guaranteed users. Mr Fossey said that papers were being prepared for the various boards' secretariats to inform them of just this kind of detail.
21. Mr Fossey said that past experience had shown that the guaranteed users took about 60% of their allocations overall. However, some of the guaranteed users used not only their full allocation but took substantial amounts of bonus time. It would be possible to over-commit the time to ensure that less bonus time were available. It was likely that such a practice would prevent the Laboratory from providing time on a guaranteed footing and would inevitably lead to dissatisfaction among users.
22. The Chairman said that there was a dilemma. The case for extra computing facilities would be clearer if the Committees and users could demonstrate the need for the facilities. However if nothing were done until the clamour arose there would be a substantial delay before any improved facilities could be obtained and brought into operation. He suggested that the Science Board be invited to take note that while, for the moment the allocation of time seemed to be under control, there were signs that significant delays would have to be introduced into the allocations in mid 1975 to meet the many demands. Provision of extra time on the 370/195 could only be a temporary solution and provision of a new machine would involve delays.
23. Professor Bransden said that in his experience the Physics Committee and Chemistry Committee had pursued a cautious policy on their allocation of time, because of their fear of over-committing the resources. Some projects had deliberately been restricted for this reason.
24. Mr Edmonds suggested that it might be too early to distinguish between the growth in demand due to projects already afoot and that arising from new projects. It was suggested that a period of three years would be necessary to highlight trends due to new projects.
25. Professor MacLellan said he thought it was reasonable to expect continued growth in demand in the Engineering Board.
26. The Chairman said he felt that the Committee should now alert the Science Board to a potentially awkward situation in the demand for computing facilities. The Board might wish to advise the Committees to adopt a more stringent policy on allocation of time and to indicate the line of their priorities.
27. Mr Fossey reported that Professor Burke had indicated his view that the SRC was already short of computing capacity for computational science. He felt that there was an urgent need to install extra computing capacity by 1976 and he suggested that a small body be set up to examine which computing facility of about IBM 370/195 potential was most suitable.
28. The Chairman said that the drive for increased facilities must come from the users through the SRC Committees, but the Committee would naturally steer and control the pressure. The Committee might suggest in reporting to the Science Board on the build up of demand and future commitments that a small body be set up to study the type of usage needed and the sources of the greatest pressure for enhanced facilities and to recommend equipment and measures needed to meet the needs.
29. Dr Howlett emphasised that the financial estimates represented the bid of the Laboratory. He said that the amount of money detailed for equipment maintenance might seem high. The cost of maintaining the 1906A and the PDP15 and FR80 machines was in line with their capital costs. That for data preparation machinery did in fact seem high: however, there was now a tendency away from the use of punch cards and it was likely that this sum would in consequence be reduced in future.
30. Dr Thomas said that there was no mention of income in the paper and asked if money was obtained from other government departments for computing facilities. Dr Howlett confirmed that there was a revenue from this source.
31. Dr Thomas asked if the costs for time were based on capital or on real costs. Dr Howlett assured him they were real costs.
32. The Chairman said he thought that it was interesting to see provision for new buildings which clearly must be very tentative. Dr Howlett said that the "Meeting House" role of the Laboratory could well develop in such a manner that there might be a need to accommodate more people to support this activity.
33. Dr M Elder was introduced to the meeting for this item. Dr Howlett said that it might be a bit early to introduce this proposal but it might help SRC to have notice of it. Mrs Paton said that if the case could be substantiated funds might be available earlier.
34. Dr Elder said that the urgency of the need for a second microdensitometer would depend on the number of X-Ray users. At the present time five packs of films were awaiting processing on the microdensitometer and would be completed by the end of November. Other packs of films would come in during that time and the delay might still remain at about six weeks. There would however be no room for other users on the equipment.
35. In response to a question about growth in demand, Dr Elder said that SRC policy in discouraging the purchase of diffractometers could be crucial. In addition, once the solution of about 20 crystal structures by the microdensitometer had been published, the microdensitometer would be seen as a practical alternative to a diffractometer. He thought many chemists had crystal structures to solve for which crystallographers would be unwilling to provide facilities on diffractometers.
36. Mr Taylor asked if the main impetus for the microdensitometer came from chemistry departments. Mrs Paton said she felt that there was no question of a second microdensitometer solely for the use of chemists. She said that the Science Board would be considering the proposal next week, would naturally wish to have the views of the Chemistry Committee on the second machine, but would also wish to consider the potential demand in other areas. Moreover, providing a second microdensitometer at the Laboratory would not significantly affect its overheads.
37. Dr Thomas said that there was a similar piece of equipment installed in the Chemistry Department at Edinburgh University as a result of funds provided by the Chemistry Committee. He believed that further submissions from the department to the Chemistry Committee were likely.
38. The Committee supported the proposal but accepted that the necessity for it must depend on the policy of the Science Board in relation to central facilities.
39. Dr Howlett introduced the paper seeking two main enhancements, an extra 32Kbytes of core store and a disc store to provide 4.8 Mbytes of storage. He said that the Laboratory was already finding the original size of the front end processor very restrictive. This was acceptable during the development stages of the project but once the machine was in regular usage as a node processor the lack of store would become an embarrassment. The whole proposal to enhance the machine was concerned with its long term development as a nodal device.
40. Dr Thomas said that the Committee would recall that both he and Mr Davies # had had reservations about the scale of the initial system. They had wanted to know the number of interfaces to be provided in the 4080 system and had wondered about its capability for extension. Although he would recommend the Committee to accept this proposal he would nevertheless like to know exactly what kind of role was envisaged for the 4080 processor. He was unclear of the objectives envisaged in the nodal processor project.
41. Mr Seddon said that NERC regarded the GEC 4080 project as a most important development in the Laboratory, particularly with regard to provision of computing facilities to the Institute of Geological Science. There were distinct advantages for NERC to gain access to the Chilton site so that the Atlas facilities and the facilities at the Institute of Hydrology could be made available to the NERC Institutes.
42. The Chairman said that the Committee recommended the expenditure detail in the paper on the enhancements proposed for the GEC 4080 front end processor.
43. Dr Howlett said that the Laboratory had been trying to provide remote job entry facilities through the Post Office switched public network. It had been found that the dial up facilities in general only provided operation at lowish speeds at for example 600 bauds. The proposal before the Committee was an attempt to rectify a deficiency in the service provided by the Post Office.
44. Dr Howlett said that recently some very sophisticated modems had been tried between RL and MSSL at Dorking. These modems had the feature of continuous monitoring of the Post Office line and had reliably transmitted data at full line speed. Since the provision of facilities by dial up on the switched public telephone network provided the great advantage of flexibility he was asking the Committee to support the case for the provision of three modems.
45. Dr Thomas said that he was concerned about two matters. Firstly, he felt that there was the risk that Universities with Post Office modems might not be able to communicate with the Chilton site. Secondly, he felt that the proposal was in effect letting the Post Office out of a difficulty which it was their duty to resolve.
46. Mr Taylor thought that the Post Office should be aware of the problems being experienced in data transmission by the Laboratory. Mr Denning said that the experience at ACL was in conflict with that obtained from many University sites. The Chairman said that it was clear that the Laboratory was not getting the services it required from the Post Office. It might be appropriate for him to write on behalf of the Committee to Mr Merriman of the Post Office Board.
47. Dr Thomas said that on Monday 21 October there would be a meeting at NEDC with the Post Office to discuss Post Office plans for data transmission. He would in fact be going himself and would be willing to represent the views of the Committee on the Laboratory's current problems on that occasion.
48. The Chairman said that in view of this the Committee would not at this stage support the request for the three modems. It would await the result of the representations of Dr Thomas on behalf of the Committee at the meeting at NRDC. Dr Thomas agreed to write to the Chairman and the Director of the Laboratory following that meeting. The Chairman would then decide whether or not to write to Mr Merriman of the Post Office. If it proved necessary, the request to acquire the modems would be decided by the Committee in correspondence.
49. Dr Howlett said that the proposals were of an extremely modest scale. The interactive graphics system was now working and a manual had been produced and it would be publicised.
50. The Committee noted that the enhancement costing less than £2000 was within the Director's authority for approval, and suggested that he should proceed with it.
51. Dr Howlett said that the Chairman of the Computing Resources Panel had asked him and the Director of the Daresbury Laboratory to write a paper on the feasibility and consequences of the relocation of the Atlas Laboratory at Daresbury. As a result, he and members of his senior staff had visited Daresbury to discuss the paper and a wide measure of agreement had been achieved in outlook, but some differences of emphasis still remained. The findings as to the practicability or otherwise of relocation had been inconclusive. The Laboratories held differing views about the numbers of staff saved by the relocation, but it was agreed that a saving of between 20 and 50 people might be possible.
52. Dr Howlett said that since the Daresbury Laboratory would be run as it is now until the closure of NINA in 1977, the proposal to relocate the Laboratory at Daresbury would have to be seen as applicable to the 1980's and to the solution of problems in computational science of that period.
53. Dr Howlett said that the covering paper referred to a meeting of the Computing Resources Panel where the lines of discussion had changed. The next meeting of the Panel had been fixed for 5 November 1974 and the three Directors had been charged with providing information on four main questions. Some of these were of a purely factual nature while others were clearly more problematical and he sought the Committee's help in dealing with them.
54. As background Dr Howlett said that he had provided some information about things which had happened in the past two or three years. It was apparent that there was an ever increasing need for the kind of facilities provided by the Atlas Laboratory which were not available elsewhere. He could not help reiterating that the computer was probably the most versatile device man had produced. The special nature of the role of the computer in forwarding computational science suggested that the Laboratory was the only place in the country where a national facility could be developed.
55. The Chairman said that the central question was whether there was a future need for an establishment like the Laboratory. He proposed to tackle the question in three ways: firstly to examine what support there might be within SRC for contributing resources to providing a central computational facility; secondly, looking outside SRC for support from the other Research Councils and the Department of Industry for the continuing existence of the Laboratory; and thirdly to examine the physical circumstances surrounding the location and equipment of the Laboratory. If the result of the discussion showed that the location could be moved to Daresbury, the technicalities, however favourable the circumstances were, would have to be explored with regard to the disadvantages of time scale and the need for new equipment.
56. Professor Bransden said that he was in some difficulty because the Physics and Chemistry Committees of which he had some understanding had been concerned with the distribution of computer resources on a fairly short timescale. While within the Universities it was clear that the Computer Board would not be able to provide sufficient computing facilities for the furtherance of some scientific projects, a considered view of the needs in the 1980's in contrast to the present ones was very difficult to obtain. His joint paper with Professor Burke presented at the meeting in May 1974 had been based upon the recognition of two separate kinds of computing in the Universities. One was of a generalised nature with moderate and sometimes large demands covering the wide spectrum of University work. The other was the large scale computation arising in projects in computational science where the scientist could only make a significant contribution through access to computing facilities of world class. It did not seem that the Computer Board was the right body to provide facilities of this category and he in common with Professor Burke would conclude that there was a strong case for SRC to provide the requisite facilities for computational science through a central computing laboratory.
57. Professor MacLellan said that the Engineering Board had set up a Panel under the Chairmanship of Professor Rosenbrock to examine the computational needs in the engineering field. Dr Howlett said that as yet the conclusions of the Panel's studies were still rather tentative. There was however a clear need for some sort of national interactive computing facility. Miss Bowell said that a survey of the engineering departments had shown that engineers were generally not receiving adequate services from the central computing facilities in their universities especially with regard to fully interactive work. There were several fields where interactive working could be developed but it was unclear whether a single central facility would be adequate. She said the survey had also shown that there was a general need for better software.
58. Professor MacLellan also mentioned that the Polytechnics would increasingly need computational support. Many were equipped with quite small computers from funds not provided by the Computer Board and had some problems of comparable scale with those being undertaken in Universities.
59. Mr Fossey reported that he had had a conversation with Professor Burke that morning about the merits of the provision of a single computer centre which had been the basis of his joint paper with Professor Bransden. He felt that the need for a reduction in the number of major computing centres in the SRC was still very clear. He recognised that the needs of the nuclear physics community were of a special nature. As a result two centres, rather than the one proposed in his earlier paper, might be the best overall solution.
60. Professor Barron said that his contact with the University world showed that there was a demand for, and a continuing need for, world class computational facilities. He emphasized the position of the Laboratory in providing unique facilities in the field of computer graphics. He also drew attention to the utter dismay felt in the University world that the Atlas Laboratory should suffer irreparable damage as a result of being move to Daresbury, mainly because there was a general fear that the Laboratory would lose the independence which characterised its present good relations with Universities.
61. Dr Thomas drew attention to the lack of investment made by SRC in computing facilities at the Atlas Laboratory in the 2 years that the reconstituted ACC had been in existence. He felt that if the special position of the Laboratory was to be maintained a considerable investment in its computing facilities was a matter of urgency. Professor Bransden said that while the discussions about the future of the Laboratory and its location had been going on, its position relative to general provision of University computing made by the Computer Board had been impaired.
62. The Chairman said that it appeared from the discussion that computing problems in Universities were of two kinds, small computing of a fairly general nature and computation central to certain scientific subjects which depended for their very existence upon large amounts of computing and upon facilities not in daily competition with the diversity required of the Universities' central facilities. The discussion had also shown that the provision of computing facilities of a special nature had not been reflected in the scale of investment made by SRC to preserve the special position of the Laboratory.
63. Mr Seddon said that NERC neither had at its disposal, nor included in its research grants to Universities, funds for the provision of computing facilities. If NERC did not continue to receive support from ACL, it would be faced, as other Research Councils and UGC might, with the need to provide special facilities for some of the research undertaken in Universities through grant aided programmes. NERC policy for the provision of computing facilities for its institutes was now closely tied to the Atlas Computer Laboratory: it would remain so for the future. Any detraction from the present service and facilities would put NERC into grave difficulty. The dependence of NERC upon ACL was increasing, although the growth was restricted somewhat by difficulties with high speed communication.
64. Mr Seddon said that the functions of the Laboratory, if relocated at Daresbury, should clearly continue to be executed as well as at present. It was essential that the service should be independent. NERC had already experienced great difficulty in ensuring that underutilised facilities within the NERC institutes were made use of by others needing that type of facility.
65. Mr Taylor said that in the Department of Industry it was accepted that ACL occupied a position unique in the United Kingdom. Its existence was beneficial to the scientific computing scene. Although the Department of Industry would wish to remain neutral on the question of location, it was worried that the prestige, expertise, and independence of the Laboratory might be damaged by such a move. However it recognised that the move might not necessarily erode the particular strengths of the Laboratory which the Department wished to see preserved.
66. Professor MacLellan suggested that a national computing centre perhaps providing facilities for computer aided design could have important implications for international collaboration in research in this or other fields.
67. The Chairman said that he had been much impressed by the seriousness of the discussion and suggested that it would be helpful if he were to write to the Chairman of SRC. He proposed to prepare a draft for discussion later in the meeting.
68. The Chairman said that on the question of the physical circumstances of location and equipment, the paper had uncovered a clear saving in manpower and that this must be kept in mind. He felt that it was not correct to assess this saving against the total budget of SRC. Nevertheless, if the move were to compromise the independence of the Laboratory the Committee might feel that it was too dearly purchased through the savings achieved.
69. Professor Bransden drew attention to the importance of adequate buildings to house the Laboratory at Daresbury. He also said that the manpower savings might not be as great as the papers suggested, since no attempt had been made to assess separately the manpower saving consequent upon the reduction in computing requirement at Daresbury following the closure of NINA.
70. Mr Fossey reported that Professor Burke had also been worried about the adequacy of the accommodation at Daresbury. He had also drawn attention to the difficulty of preserving the present independence of the Laboratory without a satisfactory solution to the management of ACL within the Daresbury complex.
71. The Committee considered the Chairman's draft of the letter addressed to the Chairman of SRC. Minor modifications were agreed and the Committee fully endorsed the sentiments expressed in the letter.
To: Professor S. F. Edwards, FRS, Chairman, Science Research Council
Dear Chairman,
After the meeting of the Atlas Computer Committee held on 16th October I feel that I must communicate to you the committee's deep concern for the future of the Laboratory. I think it will be best if I follow a logical rather than a chronological order our discussion.
2. We are all, of course, agreed that the need for an Atlas Computer Laboratory must be established by the users rather than by the Laboratory. By its constitution the ACC had a sufficient representation both from the subject committee within SRC and from outside SRC to communicate user feelings quite clearly. The background of all our discussions must, of course, be the increasing provision of computer facilities to universities and research institutions throughout the Computer Board. In the light of this development it is clear, as has been said for several years, that the future role of ACL cannot be the same as when it was originally established. The large majority of the enormous number of problems that arise in virtually every facet of university research activity will be dealt with by the computer located at universities or shared by groups of them. Naturally the demand will increase and there is every reason to expect that this overall demand for the multivarious tasks that arise in many universities will be more or les met by university facilities. However, there are considerable pieces of the research subject of many of the science and engineering departments which wholly depend on getting massive amounts of computing done competently and expeditiously. This requires dedicated facilities of a kind it is almost impossible to provide within the university environment and where a national facility is the only feasible solution. Lacking such a facility would simply imply that these subjects could not be pursued in the United Kingdom with any hope of attaining international standards. As far as can be foreseen an institution like ACL will be needed to fulfil these needs and no other way of satisfying can readily be appreciated. I want to stress very strongly that this was the clear common feature of the reports from members of this Committee familiar with grass roots feeling in universities as expressed in the committees and boards within SRC, and as personally experienced.
3. As regards other users we have, as you know, a representative of NERC who stressed forcibly the essential and probably irreplaceable part that ACL plays in NERC's programmes and potentialities. Reference was also made to the likely growth of demand in polytechnics, a demand that it was difficult to quantify in a reasonable forecast but the satisfaction of which did not form part of the Computer Board's task.
4. A national facility has to be supported on a scale commensurate with its task. The scene is set by the provision of facilities for universities and regional centres through the Computer Board. The rate of expenditure by the Computer Board runs now at many millions of pounds per year. Naturally this is distributed, rather unevenly, through the many university institutions in the country, but quite clearly a national facility could never fulfill its purpose if its scale of provision were not markedly superior to the average provision made by the Computer Board to universities. I want to stress the deep and universal concern felt in the Committee with the delay that has occurred in focusing on the future hardware needs of this Laboratory; a delay caused very understandably by many factor, including that of the future location of the Laboratory, but a delay that is hard to endure in a world that is not standing still. Future provision for this Laboratory began to be discussed three years ago and in this period the facilities at universities have been increased by tens of millions of pounds. It is urgently necessary to come to decisions on what the future facilities of this Laboratory should be and to reserve money firmly for this purpose. Otherwise the ability of this Laboratory to fulfill the needs of the particular subjects in question will be compromised because a diffusion of effort in these fields will be disadvantageous to every user and such a diffusion could well be forced in the not to distant future by an insufficiency of provision at this Laboratory.
5. Another point on which views were expressed extremely forcibly and absolutely unanimously was the need for the independence of the Laboratory to be safeguarded. This independence is essential not only in fact but in appearance. We have to work in the world as it is and the reluctance of potential users to make use of facilities whose independence is not manifest is a point repeatedly and forcibly made. in particular, a facility dominated by one type of user could not fulfil its task because many other potential users would in fact be deterred.
6. In coming now to the question of the future location of the Laboratory I must stress that although this is an issue on which many people feel deeply it is not in the same class of importance as the independence, real and patent, of the Laboratory and its possession of adequate capital facilities. The issue of location is therefore strictly a secondary issue and the advantages and disadvantages of the move can only be appreciated in the context of whether a move would delay or accelerate the provision of capital facilities and would compromise or not compromise the patent independence of the Laboratory.
7. What is certainly important is that our resources, human and material, should not be scattered. Much stress has been put in the past and continues to be put on the desirability of SRC having as few computer centres as is possible. A single one while optimal from this point of view is unlikely to be able to demonstrate its independence, as it can hardly avoid having one dominant type of user. Two centres is therefore likely to be the best solution practicable. More than two establishments is regarded by the Committee here as so unwise as to put in question the justification of a national centre at all. This question of dispersion of effort is, like independence, an issue that must be paramount in the discussion of location.
8. recent papers by Daresbury and ACL have worked hard to identify the savings and costs that would arise from a move of ACL to Daresbury. There is no question in our minds that a saving of 30-50 posts would be a very significant matter in the scale that we are discussing and is a factor that must weigh heavily in the balance. However, we feel that two questions have not so far been looked at in any detail:
9. We do not feel qualified in the absence of a. and b. above to pronounce formally on the manpower savings that would or would not result from a move to Daresbury. The feeling has also been expressed that the difficulty of fitting into accommodation at Daresbury, not purpose built for ACL, would have serious disadvantages that could be mitigated only in due course by an appropriate programme of adaption.
10. To sum up, there is clearly conveyed evidence of the need for a national facility owned by SRC. This facility must be hardware equipped so as to compare reasonably with provision at universities and regional centres and possess the necessary human resources to fulfill the need of the particular fields of research requiring such dedicated facilities. These fields are likely to wither in the absence of such a facility. To fulfill this task the Laboratory needs real and manifest independence. the question of location is of prime importance only in so far as it relates to these interests. If material savings would result the move would be acceptable in spite of unavoidable disruption. What is not acceptable is that consideration of the move could delay decisions on future provision in the Laboratory or compromise its independence.
Professor Herman Bondi
Chairman, ACC and
Chief Scientific Adviser
Ministry of Defence
18 October 1974
Minute 22 of the sixth meeting held on 23 May 1974 recorded the setting up of a panel called the Computer Resources Panel under the chairmanship of the Chairman of Council to study the feasibility, economics and consequences of the various proposals for changes in the arrangements for computing in SRC. At the first meeting of this Panel, on 24 July, Professor Ashmore and I were charged with writing a paper discussing these possibilities:
The paper which we wrote is reproduced here as Appendix I.
The second meeting of the Panel was on 6 September. There was very little discussion of the paper; instead, as the outcome of a more general discussion, the three Directors (of the Atlas, Daresbury and Rutherford Laboratories) were charged with providing information as follows:
This was required for the basis of the answer to the question, broader than that originally put on the re-grouping possibilities: what would be the effect of quantified cuts in expenditure on the computing resources of the Council on the service which those resources provide, both within the Council and to Universities? The answer to this question should take into consideration not only the present-day users of the Council's resources but also what potential users might be expected to exist in, say, five years' time.
At the time of writing the date for the next meeting of this Panel has not been fixed.
There is certain to be a great deal of serious discussion of the re-grouping proposals and, where these concern computing, this is inextricably bound up with the future policy for the Atlas Laboratory. To help me in whatever part I have to play in these discussions, it would be most valuable if I could have the Committee's views on the paper put to the Computing Resources Panel, and on the lines along which the Laboratory should be developed. If the Committee were willing to discuss this second broader issue I would offer the following as background:
All this adds up to saying that there is unlimited scope and a real need for a central professional computing institute which is not tied to a single field of application; it will not surprise the Committee if I say that the Atlas Laboratory is the one place in Britain which approximates to this, and which could be developed in this role with great benefit to science and technology.
This is a joint paper by the Directors of the Atlas and Daresbury Laboratories. Part I has been written primarily by the former and Part II by the latter. A good measure of agreement has been achieved between the senior personnel of the two Laboratories after several meetings and discussions, and although there are naturally some different points of view at the two locations, there is no wide divergence on the conclusions which were reached and are given in this paper. Where there are differences these are indicated in the text.
SRC continues to support a professional institute of the highest standard having the same role as ACL, possibly developed and extended. Its obligations include providing computational services to other SRC laboratories, without implying that these do not have their own local computing resources. The ACL view is that it is not controlled by anyone of these laboratories: it may draw on another laboratory for some services, as does the ACL on the RL, but otherwise has its own budget and staff complement. The Daresbury view is that if this institute were located there, whilst it would have a large measure of independence it would be closely integrated into the Daresbury laboratory.
The point at issue is whether this shall be at Chilton (essentially, the present ACL) or on the Daresbury site. If the latter it would be formed by moving ACL staff and some of its equipment to Daresbury, and transferring to it the present Daresbury computing staff and equipment. It would be responsible for all the computational needs of the Daresbury site, as well as the general ACL function. If it remained at Chilton it could provide some services to Daresbury but some local computing resources would have to be retained there. The question is, what are the relative economics of the two choices?
The ACL Forward Look bid gives the Laboratory's view of what is needed to fulfil its role, with the high standard required of a national Laboratory which is judged by world standards. These estimates of needs are based on attempts to forecast the demands of science in the 1980's: several papers on this subject have been put to the Atlas Computer Committee over the past 3 years and the SRC publication "Computational Chemistry and Physics" (July 1974) is relevant.
Staff: 160 (detailed composition can be provided). This number assumes that certain administration and engineering services are provided from elsewhere, at present by the Rutherford Laboratory to a small extent, AERE Harwell.
Equipment: Main-frame machine power roughly equivalent to that of one IBM 360/195 or CDC 7600, with provision for large-scale data processing and vector/array-processing; extensive ancillaries including graphics and animation; extensive and flexible communications. Flexibility is most important throughout.
Accommodation: Needed for staff, visiting users, all types of meetings, and equipment. The Laboratory view is expressed by the ACL building, which has been designed for the purpose: this provides about 4400 sq metres floor area (5164 sq metres total, including plant rooms, stores and garages), 80 offices, library, lecture theatre, common room, fully air-conditioned rooms for 2 large computers and ancillary equipment. The machine area allows quite exceptional ease and flexibility in choice, installation and modification of equipment. With future development in mind, mainly an increase in the number of visiting users (the "Meeting House" function), the Forward Look includes a new office wing to accommodate about 30 people. The present value of the existing building is about two million pounds.
In comparison with the Chilton location, which implies retaining a local computing resource at Daresbury, there could be savings on:
Consider these in turn:
Staff: The institute would need more staff to meet the special local needs of the Daresbury site in addition to the ACL function than for the latter alone: not many, but probably at least 5. The present Daresbury computing staff is 26, so the saving would be 26-5 = 21 : in round numbers, 20. The Computer Systems and Electronics (CSE) Division now includes 40 staff for electronics work, eg designing and making interfaces between the computing and experimental equipment. There will be a continuing need for these people to support NSF and SRS work after NINA has been closed and they must be provided for somewhere in the Laboratory complement. The ACL view is that they do not belong properly to the computing institute and should not be counted against the computing complement. Two people with electronics expertise are employed at ACL - mainly on communications work - and possibly 1 or 2 more might be needed at Daresbury; but no significantly greater number. The Daresbury view inclines more to incorporating the electronics staff into the computing organisation; if this were done the formal complement would have to be increased.
On Administration, the ACL view is that 10 posts could be given up in return for services provided centrally: these are 4 cleaners, 2 drivers, 2 stores and 2 library. It is to be expected that the central administration would provide also the services such as bill paying and personnel which ACL now gets from the Rutherford Laboratory, and which are reckoned to be equivalent to 10 staff. ACL considers that the remaining 16 members of the local administration group at Chilton form an integral part of the laboratory and could not be replaced by centrally-provided services; Daresbury's view is the opposite.
Thus in the ACL view the maximum possible staff saving is 30, made up of 20 on computing and 10 on administrative services; and that this will be realised in practice only if the Daresbury administrative organisation can provide the services represented by the 20 staff above (10 from ACL, 10 equivalent from RL) without increasing its own complement. The Daresbury view is that more can be accommodated within their present Administration and that a total staff saving of about 50 can be made.
Most of the staff saved in this process would be in the lower-paid groups. Taking the gross annual cost per head as about £3,000, the equivalent money value is between about £100,000 a year on the ACL view and about £150,000 on the Daresbury view.
Accommodation: A requirement of great importance is that the staff of the computing institute should all be together in coherent accommodation, close to the equipment. This has been the case at Chilton and has contributed a great deal both to the efficient operation of ACL and to its development into a distinguishable, and distinguished, professional institute. Daresbury, as it is, clearly cannot provide anything comparable with the Chilton building; however, a count of rooms and floor areas shows that it would be physically possible to house the staff of the computing institute there if the computer block, a large part of B-block (made up of offices and laboratories) and some contiguous rooms were made available. The Daresbury view, based on present estimates of the needs of NSF and SRS, is that after NINA has been closed down and the HEP work completed it will be possible to release sufficient accommodation in B-block and the contiguous buildings (together of course with the computer building) to accommodate the staff of the institute. It seems likely that some building work would have to be done to adapt some of the accommodation, but a detailed study, based on firm staff estimates, is needed to settle this. Any future expansion would almost certainly entail significant building work.
The Daresbury computer room now houses the IBM 370/]65 system and the communications equipment. It would be possible to get a twin-processor 370/168 into the room, provided that the part which now has only a shallow false floor were excavated to provide a deeper under-floor space. It would not be possible to get two independent systems, each with its own peripherals, into the room; even with the twin 168 and peripherals on the scale which would be needed, conditions would be decidedly cramped.
It seems that room for the main ancillary equipment, such as the FR-80 microfilm recorder, the graphics system and the microdensitometer could be found in the various laboratories contiguous to the computer block and B-block. There would be some costs incurred in adapting these, but again a detailed study would be needed to assess these.
Assuming that no extra costs were incurred beyond what Daresbury already pays in the upkeep of these buildings, and that the ACL buildings were not used for any purpose which SRC already pays, the saving here would be the cost of maintaining the ACL building. This is about £25,000 per year.
Equipment: The dominant question here is that of the main computer. We assume that the ICL 1906A would not be moved. Certainly it could not be housed in the Daresbury machine room. As there is no possibility of a physical move before 1979 at the earliest, we are discussing what is needed to support science in the 1980's.
Daresbury estimates that the computing and data-processing needs of the NSF and SRS are equivalent to about one-half of a full-time 370/165; if the support for CERN users continues the total load will take practically the whole time. Measurements have shown that for the ACL work-load the Daresbury 370/165 is equivalent to between 30% and 40% of the Rutherford Laboratory 360/195; and a 370/168 to about 50% of a 360/195. Thus if the ACL estimate of future needs (see paragraph 4) is accepted as being anywhere near what will turn out to be the truth (and all past estimates have turned out to be far too low) then a single 370/168 at Daresbury (and a fortiori a 370/165) is completely inadequate to meet the combined future need and even a twin-processor system is insufficient.
The question thus arises of a completely new system. It arises also, and independently, because of the time scale: if no move can be made until 1979 it is far from certain that one would then be considering buying a new 370/168. It seems likely that IBM will be offering a new range around that time (the initials FS, meaning Future System, are already being heard), that this will have some compatibility with the 360/370 series and that the needs of the experimental and data-collecting equipment will exert very strong pressures in favour of staying with IBM. But the physical needs and the problems of change-over are quite unknown now, and the questions of suitability of the Daresbury accommodation and cost of suitable new accommodation must remain open.
Thus we have reached these conclusions concerning the choice of Daresbury for the SRC computing institute.
The ACL view of the conclusions thus reached is this:
In this section the implications of retaining the Atlas function at the present site are examined.
The first important question is whether or not the Daresbury 370/165 should be moved to the Atlas Laboratory, and it can be answered by considering the probable workload over the next five years of the relevant Daresbury programme. The present plans are that the NINA experimental programme will continue until 1977, so, taking into account post-experiment analysis, it is clear that the 370/165 should not be moved until at least 1978/79. In the last years of this decade the minimum load on Daresbury computational facilities will come from NSF, SRF and theoretical computations, and these are estimated to require at least 0.5X 370/165, and much of this load will be generated on-line. Since the cost of providing this order of computational power in say 1980, by purchase of a new system, is estimated to greatly exceed the then second-hand value of the 370/165, and since such a change would inevitably cause disruption in services, it clearly makes no operational or economic sense to transfer the computer system.
Quite independent of the above, it can be argued that it is just not relevant in 1980 to base any Atlas Laboratory on a second-hand 370/165 which, by that time, will not merely be an outmoded machine but will in any case not provide sufficient capability to meet the Atlas needs.
On the question of staff numbers, it is not practicable to provide an adequate service to NSF + SRF + Theory, using collectively about 0.5 x 370/165, with fewer than the present number of computer staff in the CSE Division (about 30). Whether or not there is a change of computing system at Daresbury, no staff economies can be made. Bearing in mind the significant and continuing on-line component, it is advantageous to continue the present combination into one division of computing systems and electronics.
There will certainly be an effect on the quality of staff and services provided at Daresbury in the domain of computing and instrumentation if the Laboratory programme consists only of NSF, SRF and Theory, while both the Rutherford Laboratory and Atlas exist with greatly superior computational facilities and more challenging computational problems. It can fairly be said that it will not be possible to attract or to retain staff of the highest quality in the relevant areas at Daresbury. Bearing in mind that there will be some spare computer capacity under this model, one solution would be to retain a HEP (CERN) component. Another would be to add some other similarly demanding element to the Laboratory programme.
Notwithstanding the situation in 1980, there will be a global computing problem within the SRC and it is certain that there is now inadequate computing power to meet all the coming needs of the Council. The Daresbury view is that more computing power can be provided economically and expediently by first replacing the present 370/165 main frame by a 370/168 and then adding a second 168 in a dual processor arrangement. The ACL view is that the 370/165 will easily meet all the essential on-site needs at Daresbury for many years to come and should be left there and dedicated to that purpose: any future large computer should be installed at Chilton and that Daresbury and other SRC laboratories should be given access to it by means of data links.
