Jump Over Left Menu
Minutes of the meeting held on 17/ 11/1967 at State House, London
- Lord Halsbury (Chair)
- Dr R A Buckingham
- Professor T Kilburn
- Mr F J M Laver
- Professor E S Page
- Professor R E Peierls
- Dr T G Pickavance
- Professor M V Wilkes
- Professor B H Flowers (Items ACC/67/6 and ACC/67/7)
- Dr J Howlett
- Mr C Jolliffe
- Mr R St J Walker
- Dr A H Halsey
1 Minutes of the Last Meeting
Professor Flowers attended in his capacity as Chairman both of the Science Research Council and of the Computer Board.
The Committee approved the Minutes of the meeting held on 17th July 1967.
2 Matters Arising
- Users' Committee Dr Howlett told the meeting that he now had the names of the university representatives: Dr L A G Dresel of Reading and Professor M H Rogers of Bristol. He would arrange a meeting of the reconstituted Committee during the next university term.
- London University It had proved possible to give London more time and the records showed that they had been getting an average of about seven hours a week since April. Professor Buckingham said that, whilst they could use more, this increased allowance had helped them considerably and was much appreciated.
3 Membership of the Committee
The Chairman said that it would be of help to both bodies in co-ordinating their policies if Mr L F Rutterford, Secretary of the Computer Board, were to join the Atlas Committee as an Assessor. This was agreed.
4 Progress Report - ACC/67/4
In discussing machine performance, the Committee noted the high level that had been maintained for many months. Asked about the other Atlas machines, Professor Buckingham said that the London machine was running at about the same level, perhaps a little lower; Professor Kilburn, that the Manchester machine gave about 2% lower; and professor Wilkes, that the Cambridge machine ran at a much lower level.
The Committee noted the various tables. Dr Howlett apologised for the confusion caused by using the word job in two senses: in most tables a job meant a single run on the machine but in table 6 it meant a project. Mr Walker said that Table 6 provided striking evidence of the extent to which research workers with large computing projects could plan their work for a long period ahead.
Mr Jolliffe suggested that a statement should be sent at the end of each financial year to the other research Councils, showing what computing had been done in their fields of interest, and the cost, and inviting their comments. Dr Howlett agreed to do this., adding that these Councils were charged for any work done directly for them.
Discussing staff, the Committee commented that the complement of research posts gave opportunities not found anywhere else in the country. They noted that the present holders were all concerned with applications of computers; they would like to see more emphasis on computer science itself. Dr Howlett took note of this, but said that the Laboratory had found, as everyone else has found, that really good people in this field are very rare. He noted also the Committee's concern that the Laboratory might appear to compete with the universities for scarce research staff.
5 Atlas Costs and Charges - ACC/67/5
Mr walker said that the Finance Branch accepted the general principles on which the actual cost of a year's work had been calculated, as set out on pages 1 to 3 of the paper. They differed from the paper on some points of detail, such as the way interest should be calculated, but their process and Dr Howlett's produced almost identical results. They would send a note of their method to Dr Howlett, who agreed to use this for future calculations.
The Committee agreed that the present formula was producing an economically justifiable charge and need not be altered, at any rate until a later review showed that a change was necessary. Mr Walker said that the difference between clock time and instruction-count time was confusing and suggested that the form of the charging formula should be altered to make the two scales agree. Dr Howlett said that this was easily done, but that for consistency the present form should be used until the end of the 1967/68 financial year. the Committee accepted this.
6 Policy of Charging for Atlas - ACC/67/6
The Committee discussed this paper in the context of circumstances:
- the university Regional Computer Centres, in accordance with the terms of reference of the Computer Board, will charge all users for any work done; and
- the Science Research Council does not charge university users for any of the services it provides, for example, for use of the nuclear particle accelerators or of the astronomical telescopes.
Professor Flowers said that as a general principle it was essential that the real cost of any research project should be known, and the computation was almost always a part, and often a large part, of a piece of work. He said also that a very difficult situation could arise if university workers had the choice between a free service at Chilton, and a Regional Centre service for which they must pay.
The procedure should take account of scientific and accounting considerations. It was of prime importance for universities to be able to get on with their research without the extra administrative work and the delays which would be caused if grants had to be found for many pieces of computation, and it was also important that the Science research Council should not be burdened with the work of dealing with large numbers of extra grant applications. There was a danger of computers - which had been bought with public money - being underused because of lack of money to pay for time; Professor Wilkes said this had happened in certain American universities.
The Committee agreed that the cost of all work should be known, but noted that the Atlas Laboratory already costed everything done on the machine, that every university got a weekly statement of the cost of each piece of work done for any of its members and that virtually any information on work done, and its cost, could be provided for any interested body. Professor Peirls said that the essential problem was to keep control and prevent serious waste, but that this could be done by expert scrutinization of large projects and retrospective study of the work done for departments or research groups. This was basically what was proposed in the paper. Mr Walker said that the question of authorisation was important, as the Atlas Laboratory could not pronounce on the value of a proposed piece of work in, for example, chemistry.
The Chairman invited Mr Jolliffe, Mr Walker and Dr Howlett to formulate a procedure which would ensure proper control of the work coming to the Laboratory, and be workable by both the universities and the Science Research Council. This could be along the following lines: that small projects continue to be accepted as at present, but that the advice of the chairman or other member of the relevant Science research Council committee be sought in connection with large commitments, the Laboratory reserving the right to submit very large demands for consideration by the full committee.
7 Future Development of the Atlas Laboratory - ACC/67/7
The paper had been produced in response to an instruction given to Dr Howlett at the previous meeting (ACC/67 First Meeting, paragraph 5). It made a comparison of the CDC 7600 and the ICT P.51, and proposed that an ICT P.51 system should be ordered for installation at Chilton early in 1970. The Committee discussed the paper in the light both of the conclusions it had previously reached on the function of the Laboratory, recorded in the above Minutes, and of the SRC as a whole, which was primarily to stimulate and support advanced research by providing in its laboratories facilities not available anywhere else in the country. There was some obligation to support British industry and technological development, but only so far as this did not prejudice the main aim.
The Committee agreed that the paper gave as objective a comparison of the two machines as was possible on the facts available.
The main points made in the discussion were:
- The choice would in fact be between the CDC 7600 and the ICT P.51. The 7600 was the most powerful of any known to be under development in the USA and in the UK the plans of the only other manufacturer in the field, English Electric/Elliott, were not finalised and were on a longer time-scale than those of ICT ltd. A decision in favour of the 7600 would be effectively a postponement of the decision, since no offer of price, performance or delivery had been made by CDC Inc.
- The decision of the Atlas Computer Laboratory for or against the P.51 would have an important effect on the attitude of other potential buyers of this machine, at home and abroad.
- The Laboratory would contribute considerably to the development of software for its new machine, and this would be of great value to the manufacturer.
- It was most important that the country should press ahead with development of advanced computers and not be content to lag behind American development.
- Practically-directed research in computer science, development of advanced software and the training of experts in these fields were as important a part of the work of the Atlas Computer Laboratory as was the provision of a service. This needed powerful and up-to-date machinery but not necessarily the biggest or fastest computer available anywhere.
- If there was a clear need for a big American machine (e.g. a 7600) for university work, a good case could be made for putting it in a university; partly to give universities direct experience of American technology and software, partly to let the Atlas Computer Laboratory put more of its effort into computer science, from which all universities would benefit.
After expressions of individual views by the members, the Committee unanimously supported the main proposal of the paper, namely that an ICT P.51 should be ordered for the Atlas Laboratory. Two points were left for further discussion: the provision of remote terminals and communications equipment, and the use of Atlas after the new machine was in regular operation. The need to get firm prices from ICT was stressed, as was the fact that the decision depended on the ICT project being supported by the Government.
8 Date of Next Meeting
It was agreed that the next meeting should be held at 11.00am on Monday, 11th March 1968, in the Atlas Laboratory.