The Committee approved the Minutes of the meeting held on 4th January, 1967.
Dr Howlett presented his report. When speaking of staff he mentioned the good relations which had been built up with the Carnegie Institute of technology at Pittsburgh. One of the senior programmers, Mr F R A Hopgood, was spending the present academic year there and would be replaced by Mr M R Bird for the coming year. It was hoped to exchange staff between the two centres.
3.1 Machine Performance The Committee agreed that they wished to see comparisons between the current state of the installation and that obtaining at the last meeting rather than monthly figures. The Chairman said that in particular he would like to have more information and in more detail about the rate at which work was processed by the machine, and also about the cost of this expressed both as cost per unit of work (in some agreed measure) and cost per unit of time. Such figures would give a measure of the efficiency with which the installation was running, and comparisons between successive periods would show how this efficiency was changing.
The Committee noted the smallness of the amount of time spent on compilation. An important implication of this was that there was little more to be gained by even large increases in compiling speeds. Any significant gains must come from improvements in execution speeds.
The performance figures shown in Table 2 were discussed and the high efficiencies for the past few months were noted, particularly the value of 97.6% for the four weeks ending 19th July. It seemed that this machine was running at a slightly higher efficiency than the other Atlases, but it was recognised that comparisons between different establishments were difficult to make and, because of differences in principle and practice, might have little value. The Chilton figures were, of course, available to ICT if they cared to make use of them for publicity purposes.
3.2 Costing Formula There were some discussion on whether or not the present formula was still valid. Dr Howlett said that it was not a simple matter to cost the work done on Atlas between input, output and computation could all go on simultaneously. The present formula took into account the number of instructions obeyed and the amounts of input and output for each job, and the parameters in the costing formula had been chosen to give an average charge of £205 for each hour of scheduled time. This, on a number of assumptions also agreed at the time, was the total cost of operating the Laboratory, including writing off the capital cost of the machine. It was a basic principle that the Laboratory would make charges to other government users which represented as closely as possible the actual cost of doing the work; thee was never any intention of making a profit, and the rate was much less than would be charged by ICT in their Atlas Bureau. At present the calculated value of an hour's work was averaging about £230, the increase over the predicted £205 being due to more efficient operation of the system than had been expected. The costing formula would have to be revised to take into account increases in wages and prices, and also to allow for the increased capital value of the installation when the disc store, satellite computer and microfilm plotter were added. Mr Miller said he was about to start a study of the whole costing system. Mr Jolliffe said that the Council would be interested to have each year an estimate of the notional value of the Laboratory's services to universities, and suggested that the figures should be compared year by year.
3.3 Disc Store Dr Howlett said that all the modifications to the machine in preparation for the disc had been made without causing any disturbance. The new 16K working store was in regular use. Professor Wilkes said that the Cambridge disc (also from Data Products Inc but of smaller capacity and an earlier model) had given a great deal of trouble, and that there were no sign of improvement. The most common errors were reading and writing to or from wrong areas of the disc. It was unfortunate that there was no alternative to Data Products. Professor Kilburn said that the Manchester disc (the same model as that to be delivered to Chilton) had just arrived. It had been promised for November 1966, so the company's performance on delivery was poor. So far Manchester had had no operation experience.
Dr Howlett noted the suggestion from Professor Wilkes that, if matters did not improve, someone might be sent to the Data Products factory in America as a joint representative of the Atlas Laboratory, Manchester and Cambridge.
In opening the discussion on this paper, the Chairman made the general comment that the detailed working out of the financial provision was a technical matter for the Council's Finance Officers. It was complicated by the need to phase expenditure throughout the Council so that total spending for each financial year was as close as possible to the estimates for that year; the effect of a significant underspend caused by slippage or inaccurate phasing could be to reduce the amount of money available in a later year. The Committee had to decided the major features of the Laboratory budget: Mr Miller said that once a project - for example, the addition of the disc store - had been approved, the financial consequences followed automatically and gave very little room for manoeuvre.
Dr Howlett said that the only really important new item in the 1968/69 Estimates was the provision for the new computer. An offer from ICT Limited dated 12th July for a PF 51 system, circulated at the meeting as an Addendum to ACC/67/3, gave a budgetary price of £1,546,800. The provision of £500,000 in each of the three successive beginning 1968/69 would account for the main progress payments on the system, and in fact tallied almost exactly with the sum provisionally allowed in the Forward Look. It might prove that no money would be spent on the new computer in the year 1968/69 if the contract were not completed in that financial year, and in that case the payments would stretch out over a longer period and the Forward Look would have to be revised.
Asked about the large item for equipment, supplies etc in the recurrent expenditure, Dr Howlett said that the main components of this were the maintenance of Atlas (£155,000 a year for four shift working), the charge for Rutherford Laboratory overheads (£50,000) and the Group Budgets (£80,000), of which the Operations Group accounted for £60,000. The sum for salaries, wages and superannuation (£155,000) for 1968/69 followed accurately and automatically once the complement had been agreed. The committee formally approved the Estimates at the total of £1,142,000.
Before asking Dr Howlett to present his paper the Chairman gave the meeting a background account of what had taken place in the University Science and Technology Board and the Science Research Council, and in the Computer Board, and he enlarged generally upon paragraph II of ACC/67/3. It seemed that the only suitable machines likely to become available within the relevant period would be the CDC 7600 and the ICT PF 51. Concerning the CDC 7600, it had so far proved impossible to get sufficient information from the company on which to assess this machine. The Chairman had written to Mr Norris, the President of CDC whom he knew personally, with a request for this information. ICT's former 1908 proposal had been examined by a small panel appointed by the Computer Board, and they had proposed certain modifications which had led to the PF 51 concept. At the present time the Ministry of Technology were considering whether or not to sponsor this machine. They were in close touch with ICT, who were keen on the PF 51 as a concept but had to be assured of sufficient orders before they could evaluate its commercial possibilities.
The situation was now that the University Science and technology Board had made financial provision for a new machine in the Forward Look and was inviting the Committee to decide on technical grounds which machine should be installed.
The Committee then discussed the principles underlying the proposals of the paper, and came to the following conclusions:
With regard to the time-scale of the development of the Laboratory, the Committee agreed that the present Atlas computer would become inadequate between 1970/71 and 1972/73. To provide the overlap needed to ensure smooth running of the service, a new machine should be operating by 1970/71.
The Committee considered the relative merits of the CDC and ICT projects, so far as was possible on the basis of the information available. They concluded by instructing Dr Howlett to assemble the information needed for a technical assessment, and to put forward a precise proposal at the next meeting.
It was agreed that the next meeting should be held at 11.00am on Monday 13th November 1967 in the Atlas Computer Laboratory.