Minutes of the meeting held on 30/ 9/1966 at State House, High Holborn, London

• Present
  • 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
  • Dr R F Churchhouse
  • Dr J Howlett
  • Mr C Jolliffe
  • Dr J H Price
• Secretary
• Apologies
  • Dr A H Halsey

1 A Satellite Computer and On-Line Time-Sharing System for Atlas - ACC/65/5

1. Lord Halsbury said that this was his first meeting as Chairman of the Committee, an honour he had been most pleased to accept. He had thought it proper (it being the Committee's duty to advise the University Science and Technology Board) to hold a special meeting on a proposal of such importance to the Laboratory rather than to deal with the matter by correspondence.

2. Presenting his paper Dr J Howlett stressed that, while it was not the Laboratory's intention to emulate Project MAC, on-line time-sharing was both one of the most important developments in computing and one in which there was little practical experience. So far the Cambridge Titan was the only machine where on-line time-sharing was being put into operation. Further, a great part of the Laboratory's own programming work was directed to software research which would be carried out more quickly through consoles. This would be equally true of development work by university users. There was also a number of interesting projects, (for example, the development of a language for algebraic manipulation on computers suggested as a joint project with the University of East Anglia), which could not be done without time-sharing facilities.

For the technical reasons set out in the paper the only satisfactory means of setting up such a system on Atlas was based on a satellite computer. All the important manufacturers had been approached and the choice of machine lay between a CDC 1700, a PDP9 (both American machines), an SDS Sigma 2 (an American machine being produced in the United Kingdom by GEC) or, possible, a Ferranti Argus. A final choice could be made shortly.

3. ICT Limited (who were responsible for the maintenance of Atlas and the Supervisor) were interested in the project and willing to provide some help with the software development as they were anxious to obtain information on the performance of a time-sharing system on Atlas.

4. Dr Churchhouse then reviewed the historical development of relationship between the programmer and the computer. In the early 1950s the programmer had himself operated the machine while his program ran and could, for example, correct minor programming errors at the time. The advent of more complex machines had led to running by a professional operating team which, while giving higher overall efficiency in machine use, broke the direct link with the programmer and led to wasted machine programming time (where there was, say, a small error in the program but it took some hours or even a day for the program to pass through the operational queue, return for correction, and go again to the queue). The development of multi-access systems allowed the retention, in the main, of higher efficiency given by batch processing, while apparently restoring the direct and immediate link between the programmer and the machine for development work. Dr Churchhouse had recently seen and used multi-access systems in a number of American institutions and had been very impressed, subject to proper control of the system, both by its advantages to the individual, and by the way it might open up possible new fields of work (for example, in information retrieval).

5. The development proposed at the Atlas Laboratory would go through two stages. In the first stage a job would be first handled by the satellite machine and eventually passed to the disc store; the main computer would be informed of its arrival. In this way the job would join the Atlas job queue and run in normal mode. The second stage would allow a certain percentage of Atlas time to be devoted directly to time-sharing on on-line jobs. Since it was difficult to estimate quite what effect consoles would have on batch processing efficiency, the Supervisor would hold instructions for cutting off consoles automatically if they took too much time and made serious inroads into batch processing efficiency.

6. In discussion, the following points were made:

1. The proposal was right in principle. It was a natural extension of the Atlas computer, which had certain design features making it particularly suitable for multi-access use. It was very desirable for the United Kingdom to have experience of on-line time-sharing, both from the users' and the manufacturers' points of view.
2. The Laboratory believed that the necessary software development could be undertaken by the equivalent of six programmers (which would come from the existing staff and complement), together with the help promised from ICT Limited. In this connection Professor Buckingham said that the University of London Institute of Computer Science would willingly give programming help as the Institute was very interested in the project and would like to be associated with it.
3. It was essential to recognise that, besides the formal costs of the proposal, it would inevitably reduce the batch processing time available both during development and in operation. The universities were, at present, very dependent on the Atlas Computer Laboratory for computing power and could reasonably ask for some reassurance on the effect of this proposal on the laboratory's ability to meet an increasing demand. However the Rutherford Laboratory was, at present, taking some 30-40% of available time. This time would not be taken when the Rutherford's new machine (to be delivered in November 1966) was commissioned. The loss of this major user together with the planned extension to week-end working should be sufficient to ensure the ordinary computing service could meet demand.
   Users would also gain from access to the consoles. The majority of consoles would be in the atlas Laboratory and some of these would be available to suitable visiting users.
4. A further hidden cost was involved in the devotion of a skilled programming team, at present one of the country's limited resources, to this particular project.
5. The proposals had no influence on the timing of a replacement for the present Atlas computer.

7. The Committee concluded that:

1. It should be their policy to increase and improve the Laboratory's facilities. On-line time-sharing was a valuable facility and a particularly suitable field of development for the Laboratory since, largely through lack of practical experience in service use, little was known about the most economical way to construct and control such a system.
2. The Atlas computer was a very suitable machine for an experiment in the field, having anticipated design features which were common to a number of large machines at present under construction but not delivered. Also, the Chilton Atlas still had a sufficiently long useful life to justify the development.
3. The penalty in lost batch processing time and in use of manpower was not such as to override the advantages.

8. It was agreed:

1. to recommend to the University Science and Technology Board that a satellite computer and on-line time-sharing system be provided for the Atlas Computer Laboratory at a cost not exceeding £95,000.
2. that the Chairman be authorised to proceed with the submission of the case to the Board on receiving the Laboratory's recommendation on the particular machine to be ordered. Copies of the Board paper would be circulated to the Committee and any comments from members would be reported verbally to the Board by the Chairman.

2 Specification of Design for Future Computers

Professor Kilburn drew the Committee's attention to the importance of computer users specifying their design requirements for machines five years in advance; it was not sufficient to leave the design of future machines solely to the manufacturers. The Chairman undertook to enquire of the Computer Board whether it regarded encouraging such planning between users and manufacturers as part of its duties and, if so, what mechanism it proposed to set up to this end.

3 Next Meeting