Before the formal business started Dr Howlett made two statements:
The Minutes of the Meeting held on 31 March 1971 were approved.
Official Opening of the 1906A The Committee noted that there had been a modest opening ceremony for the Oxford University 1906A, that Leeds were planning something for their machine and that the Department of Trade and Industry were intending to make something of an occasion of the opening of the Business Statistics Office machine. They reaffirmed their support for the idea of an opening ceremony for the Chilton machine, probably early in 1972, and invited Dr Howlett to make enquiries concerning possible arrangements.
The Committee noted the numerical tables.
Dr Howlett reported that the hardware of the 1906A seemed to be working well, that the George 3 operating system (Mark 5.4) had been installed and that there was some hope for starting a restricted internal service before the end of September. ICL had made new proposals for the paging hardware within the previous few days, different from those described in the paper. The change to paged store required nine new multi-layer platters and changes to about 30 existing platters; ICL now proposed to make two complete sets of these platters and to modify a machine in the field by replacing its existing platters by a new set - the old platters then being modified in the factory and used to put paging into another machine, and so on. They would do this in about eight night shifts, returning the machine to unpaged operation each morning. They expected to be able to start the operation on the Chilton machine in January 1972 and to complete it within about three weeks. If, as was to be expected, the machine was not then required for three shifts each day, this procedure should make the change to the paged form without any disturbance to the computing service. The proposal had however been made too recently for there to have been time to assess it. The Committee noted this and left Dr Howlett to consider the proposal and to negotiate with ICL.
Dr Howlett reported that all the display and graphics equipment had been received from DEC.
The November launch date for the UK-4 satellite had been confirmed. The data-processing seemed likely to need between five and ten hours a week on the 1906A. That for the S-68 experiment seemed likely to need about five hours a week on the 1906A.
Dr Howlett reported that the Laboratory had just undertaken to do the data processing for the Southern Sky Survey, a survey by optical telescope due to start in late 1972 and to take about two years. The Galaxy machine at the Royal Observatory Edinburgh would scan the photographic plates from the telescope and write the digitised information (on star positions and magnitude) on magnetic tape. The object was to construct a large data bank for use in astronomical research. The total amount of information to be stored would be between 1010 and l0ll bits, equivalent to about 1,000 reels of magnetic tape, so there would be severe problems of data organisation.
The paper was for the Committee's information and did not ask for any action.
Dr Howlett said that the original undertaking of the SRC was to finance the use of this machine for two years only and that this period had ended on 31 March 1971. They had agreed to extend the period for a year as a result of a very strong plea from the Scottish universities but were quite firm in saying that there would be no further extension beyond 31 March 1972. Mr Rutterford said he was concerned about the situation in Scotland. It now seemed unlikely that a new computer could be installed at Glasgow before late 1972 and if access to the 1108 were withdrawn a number of research workers in that area would be seriously embarrassed. Possibly some of the work could be transferred to the Rutherford Laboratory 360/195. He felt that there should be a meeting between SRC, NEL, the Computer Board and the Scottish universities to decide on a method of procedure. The Committee agreed that this would be sensible but felt that there was little which they, as the Committee, could do in this connection.
Dr Howlett said that he had nothing of any substance to report. There had been some exchanges of letters between the Atomic Energy Authority (who were acting for the Science Research Council) and the National Computing Centre, but no progress. The NCC seemed to have become reconciled to the fact that the market for the system was much smaller than they had estimated but had not come to the point of deciding to drop the project. The formal agreement had not in fact been signed. Mr Taylor confirmed that the NCC did not wish to put any more effort into ASCOP, but were still hoping to recover some of the large amount of money which they had spent on it. Dr How1ett said that the system was being used on Atlas, but that he was not able to give copies to universities. The Committee re-emphasised its view that the situation was most unsatisfactory and asked Dr Howlett to continue to press for a resolution.
The Committee noted the paper and Dr Howlett's intention to make a firm proposal for a machine early in 1972. They ratified Mr Jolliffe's provisional authorisation for an increase in the budget for the development work from £3,500 to £4,500, Mrs Paton said that because of the possibility of commercial development the paper should have been marked Commercial in Confidence.
Mr Rutterford said that the general question of operators' information on the 1906A's was being discussed by a section of the Joint 1906A's University Group, and that there would be support for the proposal from all the universities. The Committee discussed the technical points of the paper and approved the proposal at a total capital cost of £13,700 with annual maintenance charges totalling £107-10.
This was included in the discussion on future policy (see Minute 10).
Whilst recognising that Atlas was an expensive machine to maintain and operate in comparison with newer computers, the Committee were concerned that many users who relied on it might find themselves in difficulties when it was withdrawn from service. They supported the proposals of the paper - to close the machine at the end of October 1972 and to accept no new projects after 1 June 1972 - but asked Dr Howlett to find out from users how they would be affected by such a decision.
Dr Howlett was seeking the Committee's advice on the content of the paper he would put first to the Science Board at its meeting on 13 October and then, if the Board approved, to the Council on 15 December.
The Chairman said that this had been discussed and approved (eg ACC 70/8) but the Committee felt it would be helpful to note the main lines. It was agreed that the Laboratory would gradually, but as quickly as seemed reasonable, change its mode of operation into one in which, so far as university work was concerned, its resources were used mainly to support projects which had been submitted to and approved by one or other of the SRC's Boards or subject committees, and that these projects would be guaranteed time and other facilities and services. The over-riding criterion for acceptance would be the quality of the science, but given this it was also necessary that the facilities needed by any particular project could not be provided in the originator's own university. Demands for time and storage were likely to be the dominant factor, but there could be others such as the need for special software or ancilliary equipment - microfilm output, for example. The Laboratory needed to have a vigorous Research and Development programme of its own to ensure the continued quality and 'effectiveness of the computing service, and this also would be guaranteed time and other facilities.
Implementation of this po1icy would start with the 906A and 360/195, and initially at least not all the time available on these machines would be committed for approved and guaranteed projects. Some fraction should be left for allocation at the Director's discretion. The Committee did not have any rigid view on the size of this fraction or on whether or not it should be gradually reduced to zero, but felt that as the implementation of the new policy was bound to be an evolutionary process it would be sufficient to keep the situation under review.
Professor Kilburn commented that in the new regime the Laboratory would for the first time be playing a role which was essentially different from university computer services, and the Committee agreed that this was so. They agreed also that it was essential that the Laboratory should be different, since otherwise there was no justification for its existence. The Chairman hoped that by operating in this way the Laboratory would encourage research workers to formulate projects of good scientific merit which they would not even consider if they could not be assured of adequate computer time. With the Laboratory meeting at most about 5% of the total university computing load it was the type and quality of its service which justified its existence, rather than the quantity.
In response to a statement by Mr Berman that some form of financial control was needed, Mr Jolliffe said that the cost of the computing requested by any applicant would be estimated by the Laboratory and noted on the application. The committee or other body to whom it was referred would thus see the total cost of the proposal and would have to decide whether or not it merited what was in effect a grant of that size. He did not think there was any need for actual payment of money, because in these circumstances that would be simply a book transaction.
Mr Taylor, for the Department of Trade and Industry, suggested that the Laboratory might take on the task of identifying surplus capacity in existing computers and allocating this to applicants needing more time than their local machines could provide. He felt that this task called for a unit of high professional standing and ability, and that it would need not only administrative competence but also skill in advanced technological developments. He believed that the Atlas Laboratory met these requirements well and that the task which he had suggested would complement effectively any role concerned with offering a selective service. The Committee felt that this would raise severe financial problems when computers in government departments were concerned (instance the use of the National Engineering Laboratory machine), and in any case was not an appropriate role for the Laboratory.
The Chairman said that it was indeed very difficult to estimate the scale of computing power which would be needed to meet the future demands on the Laboratory, as was stated in the paper. The essential nature of the difficulty was seen when one noted that one was trying to predict the state of science seven to eight years ahead: about two years to get the machine (the view given in the paper, which was accepted), and at least five years of service. Reference was made to the two working parties chaired by Dr Francis, which were studying the internal needs of the Science Research Council and of the other Research Councils and which would attempt also to get some estimate of the part of the university need which SRC should meet. These groups would be reporting early in 1972. Dr Howlett was asked to make use of any information he could get from their current papers.
After a general discussion of the growth in demand for computing the Committee agreed unanimously that Dr Howlett's estimated range of 10-30 times Atlas was realistic, but felt that the need was likely to be nearer to the top end than the bottom. Professor Kilburn said that the demand had followed, and had been stimulated by, the developments in technology and that the end of these developments was not yet in sight; so one should expect demand to continue to increase. He said that in his view the Laboratory should install the most powerful machine which the SRC could afford; but the architecture of the machine should be considered carefully, to ensure that it was matched as well as possible to the type of work expected. In particular he felt that the CDC STAR should be studied, to find if its special vector-processing operations would give particular benefit. If the study showed this was the best machine available he would strongly recommend it for the Laboratory. On growth of demand, he recalled the fears expressed ten years ago that Atlas would never be filled and noted how quickly these fears had proved groundless. Dr Dunworth supported these views, as did the Committee generally. Mr Taylor said that the purchase of a big American machine for the Laboratory would be in direct opposition to the Government's stated policy of acquiring large computers from ICL unless strong arguments existed to debar this; his Department would clearly wish to look very carefully at such arguments. The Chairman said that whilst there could be political or national economic reasons for buying a British machine, the responsibility of the Committee was to state the scientific needs and to leave such considerations to other bodies. The feeling of the meeting was that the one British manufacturer, ICL, could not provide a machine of the power which the Committee believed was necessary. Dr Dunworth added that whilst in general he supported the policy of buying British machines, he felt that this requirement could be relaxed more readily in the case of the Atlas Laboratory than anywhere else. Mr Berman expressed some regret that, if the Laboratory were to install an American machine, American industry rather than British would benefit from the work of its programming staff. Dr Howlett said that the outcome need not be so extreme, but that a good deal of development work could be of general application.
Mr Jolliffe said that it would be necessary to convince the Science Board of the need for spending possibly about £4M on a capital project, in addition to the estimated running costs of the Laboratory which were about £O.8M a year, at a time when funds were already short and getting shorter. Some comparison with the funds being provided by SRC in support of other subjects should be made. It was agreed that the paper which Dr Howlett would put to the Science Board at its October meeting should give the general arguments for a new computer; and that a paper giving a detailed case, based on information collected on the needs for large-scale computation in specific fields or projects, should be put to the Committee in about six months t time. The Committee could then decide if it could recommend that a firm proposal should be put to the Board.
The Committee agreed that the estimates given in the paper were reasonable. In particular they agreed with views that fewer people would be needed to operate the advanced machines of the future, whilst more would be needed for software development and related work. Professor Kilburn noted that the staff numbers were very similar to those accepted as needed by large Regional Centres.
The Chairman asked Dr Howlett to proceed with his paper to the Board; because of the timing of the Board meeting (13 October, and the papers must be sent out by 5 October) there was scarcely time for Committee members to see and comment on a draft. It would however be appreciated if a draft could be sent to them, to give an opportunity for at least verbal comments over the telephone. Dr Howlett agreed to do this.
