290. In Chapters V and VI we make detailed recommendations for computer expenditure over the next five years, (tentative for the later years) in Universities, Colleges of Advanced Technology and Research Councils. These were made on the assumptions that all machines were purchased outright and that delivery dates given by the manufacturers were to be met with no slippage. Table 4 summarises these figures and it should be understood that the cost given represents the cost of hardware alone, with no allowance for extra peripherals normally running at 5 per cent per annum of basic cost, nor for buildings, maintenance, running costs, staff etc.
| 1965-66 | 1966-67 | 1967-68 | 1968-69 | 1969-70 | Total | U.S. Component |
|
|---|---|---|---|---|---|---|---|
| Universities etc. | 1.95 | 3.08 | 4.60 | 4.30 | 3.75 | 17.68 | 5.73 |
| Research Councils | 0.25 | 2.15 | 1.10 | 1.60 | 3.35 | 8.45 | 4.60 |
| Total | 2.20 | 5.23 | 5.70 | 5.90 | 7.10 | 26.13 | 10.33 |
These sums include £0.43m for a computer in the University of Belfast, which would be financed through the Government of Northern Ireland.
TABLE 4 EXPENDITURE FOR HARDWARE £M
291. In addition we anticipate two extra expenditures mostly on British equipment, namely £0.5m for teaching hospitals and £1.5m for Technical Colleges etc.
292. The purpose of the rest of this chapter is to estimate the actual rates of spend allowing for the likelihood that most of the large American computers will be hired, the likelihood that "slippage" on delivery dates and on buildings will occur in some but not all cases, and to make reasonable estimates, which must be regarded as tentative but realistic, of maintenance, building and extra staff costs.
293. It has been assumed that the annual hire charges on the large American machines is 25 per cent of the capital cost of the installation. This estimate is intended to cover also the cost of hiring all the immediately accessible peripheral equipment. The Edinburgh, Aberdeen, and University College, London, installations however would be purchased because it is hoped that they can be obtained quickly and would run for the whole period under review, in which case hiring would be an uneconomical choice.
294. It is assumed that the majority of the KDF 9 upgrading equipment will be delivered during the financial year 1966-67 and that payment for this equipment will be made on delivery. It is assumed also that it is not practicable to estimate the delivery date of a computer, particularly a computer of a new series, within six months, and therefore a "slippage" factor of one half year has been incorporated for certain machines and for building delays in order to get more realistic figures for annual spend rate.
295. Upgrading equipment for the new serial computers from April 1968 onwards has been taken as being paid for on delivery, on the grounds that by that time the industrial production capacity should be able to meet the demands on time. The resulting spend rates are shown in Table 5.
| 1965-66 | 1966-67 | 1967-68 | 1968-69 | 1969-70 | 1970-71 | |
|---|---|---|---|---|---|---|
| Universities and CATS | 1.385 | 3.161 | 2.770 | 3.164 | 3.694 | (1.761) |
| Research Councils | 0.250 | 1.060 | 1.290 | 1.875 | 1.712 | (1.200) |
| Total | 1.635 | 4.221 | 4.060 | 5.039 | 5.406 | (2.961) |
| Total Hardware Costs +5% contingency | 1.650 | 4.435 | 4.263 | 5.291 | 5.677 |
TABLE 5: PROBABLE SPEND RATE FOR COMPUTER EQUIPMENT EXPENDITURE £M
As noted in the text, these sums include a £0.43m recommendation for a computer for the Northern Ireland University of Belfast; with the 5 per cent contingency charge this will amount to £0.45m, which would be financed together with buildings, staff, etc., through the Government of Northern Ireland. This aspect was brought to our attention after our financial assessment was completed, but it was decided to leave all the figures in this report, which are internally consistent, as they stand.
296. The figures shown in the 1970-71 column are the run over values including the hire charges of the large installations and some slippage outlay for computers which should have been installed in the previous financial year. A "contingency" allowance of 5 per cent is made to cover the costs of ancillary equipment which may be required by an installation after running experience. Many of the previous requests to the U.G.C. have been for extra facilities of this kind, and it would create great difficulties for the U.G.C. and for the Universities if such a contingency were not to be held.
THE LAST LINE OF THIS TABLE IS OUR RECOMMENDED RATE OF SPEND ON HARDWARE.
297. In order to obtain estimates for the increased annual costs of maintaining University computers a figure of approximately 5 per cent of capital cost per annum has been taken. Computers on hire are of course excluded from these estimates. Where new computers are replacing obsolete machines an allowance has been made for the maintenance charges on the outgoing machine.
| 1965-66 | 1966-67 | 1967-68 | 1968-69 | 1969-70 | 1970-71 | |
|---|---|---|---|---|---|---|
| Universities and CATS | .042 | .195 | .308 | .419 | .533 | (.551) |
| Research Councils | .017 | .050 | .093 | .159 | .188 | (.217) |
| Total | .059 | .245 | .401 | .578 | .721 |
TABLE 6 INCREASE IN MAINTENANCE CHARGES EXPENDITURE £M
298. The costs included under this title include provision for the many and varied necessities of a University computer installation. The most significant operating cost in the larger installations could be for the provision of magnetic tape, and so the KDF 9's and other magnetic tape backed installations have been given special thought. The University installation would normally supply all its own stationery including cards, paper tape, and suitable storage equipment for them. The operating costs include the provision of electricity and support charges for air conditioning services where applicable. Some thought has been given to the cost of hiring telephone lines for data links, and the costs of transport of programs and results both within the University and outside.
299. For example, there will be some fifteen or twenty installations using a good deal of magnetic tape during the next five years, and an estimate of the cost of this stock together with cards and paper tape might be put eventually at £.250m. This expenditure will not be spread evenly over the five year period but will rise in much the same way as the total computing capacity increases. A similar estimate of ancillary services such as the use of telephone lines, air conditioning maintenance, and the provision of stationery stores for all Universities, Colleges of Advanced Technology and Research Councils is considered to be £300,000 per annum at the end of the five year period. A rough estimate of yearly spend might be as follows:-
| 1965-66 | 1966-67 | 1967-68 | 1968-69 | 1969-70 | 1970-71 |
|---|---|---|---|---|---|
| .050 | .160 | .250 | .350 | .500 | (.520) |
TABLE 7 INCREASE IN OPERATING COSTS EXPENDITURE £M
300. The increase of computing power projected by the proposals in this report demands that two main categories of staff, academic and non-academic, be made available. The operation of University computers, eventually over three shift working, means that additional operating staff must be provided on the larger machines in shift teams so that maximum production work can be achieved. It is important to re-emphasise that the performance of a particular computer installation depends very markedly on the quality of the operating staff; a good computer can be made much less powerful through incompetent operation, and a poor computer can be made to give much through competent operation. Depending on the size of the installation as described elsewhere in this report, a number of shift operators is required such that the shift cycle of throughput is carried out with a minimum of wasted machine time. Elsewhere in this report mention is also made of this need for professional computer operation.
301. A backing of academic staff is essential in support of even the smallest University computer installation so that the facility can be used to its fullest extent. Generally University installations run on the "open shop" principle, that is the research students and staff are able to prepare their own programs; even so the University computer laboratory should have a number of experts to advise and direct the progress of a particular project when and where necessary.
302. The principal need however is for non-graduate technical staff, i.e. for shift operators. The fifteen or twenty medium or large establishments which are intended to run extensive computer services must each plan to have a minimum of twelve non-graduate operators. This reveals a deficit at this time of approximately one hundred and fifty personnel of this type. Over the five year period in question it is considered vital that these people be employed and trained to computer operation practice in Universities. The U.G.C. is urged to consider concentrating training of these people (operators and academics) at two or three academic institutions. The Chilton Atlas Laboratory might also be able to help in this connection.
303. The need for academic and research staff is evident at establishments where medium or large computers are anticipated. Not only should these staff act as internal support for the general computing work of the University but also provide the impetus in building the output of computer trained graduates. The need is particularly acute at Nottingham, Southampton, Belfast, Strathclyde, and in almost every College of Advanced Technology. It is estimated that between 20 and 50 extra academic and research staff will be needed by the end of 1967 to support the envisaged programme. These people will be difficult to find, even though they average only about one per University, unless a conscious effort is made to train them at the existing well-established centres.
304. A cost estimate for the increased numbers of staff to support the programme might be as follows:
| 1965-66 | 1966-67 | 1967-68 | 1968-69 | 1969-70 | |
|---|---|---|---|---|---|
| Operating Staff | .010 | .040 | .060 | .200 | .300 |
| Academic and research staff | .010 | .030 | .100 | .150 | .200 |
| Totals | .020 | .070 | .160 | .350 | .500 |
TABLE 8 INCREASED COSTS OF ACADEMIC, RESEARCH AND OPERATING STAFF EXPENDITURE £M
305. These figures may seem large, but it is the view of the Group that unless the U.G.C. makes earmarked sums of this order available to Universities, the full benefit of the facilities we are recommending to them will be wasted.
306. These are particularly difficult to estimate, but our experience is that building costs for a computer are of the order of 1/6th that of the purchase cost of the hardware. We have estimated on this basis, allowing for the buildings to be installed a suitable time in advance of delivery and not charging any Sum where we know that further building is unnecessary. The figures, which are very tentative, but serve to establish scale, are given in the next section. In the case of buildings required to meet the early years of the University component of this proposed programme it is essential that the U.G.C. should note this need and be prepared to meet it.
307. We have seen that the estimated cost of all the proposed hardware, if delivered on time and purchased outright, is (rounded figures) £26.lm., of which £10.3 m represents American equipment. A more realistic estimate of money that would be actually spent during the quinquennium allows for some slippage in delivery date, for a 5 per cent "contingency charge" for ancillary equipment, the detailed need for which will come to the users only with experience, and for hiring the big American computers (the result of this being to shift a substantial sum into the subsequent quinquennium.) Under these circumstances, the total actually spent on hardware in the quinquennium would be £21.3m (of which £4.75m represents American equipment).
308. To these sums must be added £2m. estimated maintenance costs, £1.3m. operating costs, £1.1m. for the wages of extra staff at Universities to enable them to employ the full value of the equipment that will be provided for them, and £4.5m. for buildings. These sums will have to be found, by the U.G.C. and Research Councils, if our recommendations are to be adopted and utilised sensibly.
309. Thus the estimated grand totals of expenditure in the quinquennium are £35m. (if all hardware purchased outright) and £30m. (if most of the large American machines are hired). To these sums should be added £0.5m for teaching hospital hardware, £1.5m for technical college hardware, and appropriate sums for building and support facilities for these institutions. The sums include the cost of the Belfast computer: see note to Tables 4 and 5; the total cost of purchasing, housing, and operating this computer would be £0.6m. in the five-year period.
310. The Working Party urge most strongly that these sums represent the clear need of the institutions they have examined and that any attempt to reduce them will result in lack of necessary computing facilities.
311. Table 9 shows the detailed proposed year-by-year expenditure for Universities, C.A.T.s and Research Councils only, including the cost of the Belfast computer, as noted above. The sums involved are gathered from Tables 5 to 8 inclusive.
| 1965-66 | 1966-67 | 1967-68 | 1968-69 | 1969-70 | Totals | |
|---|---|---|---|---|---|---|
| Hardware, including 5 % contingency. Some machines hired | 1.650 | 4.435 | 4.263 | 5.291 | 5.677 | 21.316 |
| Maintenance charges over those already incurred | .059 | .245 | .401 | .578 | .721 | 2.004 |
| Operating costs | .050 | .160 | .250 | .350 | .500 | 1.310 |
| Increased Staff wages for Universities | .020 | .070 | .160 | .350 | .500 | 1.100 |
| Building Costs | .250 | .500 | 1.500 | 1.500 | .750 | 4.500 |
| Totals excluding Technical Colleges and Teaching Hospitals | 2.029 | 5.410 | 6.574 | 8.069 | 8.148 | 30.230 |
TABLE 9
312. In terms of a unit of computer power (one equivalent IBM 7090) we have tried to estimate how the total computing power of Universities, Research Councils and Colleges of Advanced Technology will increase if the recommendations in this report are carried out. A major difficulty which must be pointed out is that large regional machines cannot at this stage be given anything but a guessed power; since these machines are not yet available, we have had to make reasonable assumptions regarding their likely power. For the medium and smaller computers much better estimates of power are possible and the upgraded KDF 9's for example have been taken as equivalent eventually to 1.2 IBM 7090's.
313. The curve shown in figure 2 illustrates the projected increase for the quinquennium. It should be noted that the increase in power is not and cannot be obtained rapidly and that it is not until the second half of the period that rapid improvement is possible.
FIGURE 2. Computer Power Estimates 1965-1970 (Universities, Colleges
of Advanced Technology and Research Councils)
314. A growth factor of approximately 1.6 per year would result from this expenditure. This represents as we have said, the result of our estimate of clear need in individual cases, added together, and it is interesting to note how closely this growth factor corresponds to the general computer power growth factor in other advanced research projects noted earlier. Assuming that university computing services are able to run more efficiently in the future, thereby achieving an additional factor 2 in effective power over the 5 years, the annual growth rate would finally become about 1.8. It is also interesting to note how the effect of the large new machines is to enable such remarkable growth factors in computing facilities to be achieved with very much smaller rates of increase of cost.
315. It has been noted that the Group has been able to estimate clear need reasonably precisely for the first 2-3 years, and thereafter with diminishing precision. It is clear that what we have sketched out is a rolling five-year programme which must be reviewed approximately annually. There is one important point that must be made if, as we hope, our recommendations are adopted. This point arises from the 18 month-2 year waiting time which is characteristic of orders for large computers. As a result, it is important that orders for installations for the financial years 1967-68 shall be placed soon.
316. One other point of importance has emerged from our studies: the need for bulk purchasing arrangements. The Working Group is convinced that bulk orders, in numbers around 10 or more, can result in substantial savings to H.M. Government, as well as giving a heartening regularity in flow of orders to the manufacturers. It is realised that the ultimate users of the equipment should be the Universities and Research Councils concerned, but the Working Group suggest that the actual purchases should be undertaken in bulk by a suitable body and single computers resold to the users at a suitable price, thus enabling each user to enjoy the benefits of bulk purchase. Users will have to accept machines of a uniform specification in the first instance, but individual "extras" would be taken care of under the "ancillary equipment" heading.
317. The total rate-of-spend costs of the proposed 5 year new hardware programme for Universities and Research Councils established on the basis of clear need are of the order of £21m starting at less than £2m in the first year and thereafter rising significantly. In announcing this inquiry the Government anticipated that an expenditure of £2m would be incurred in the first year.
318. It is interesting to analyse the consequences of attempting to continue annual expenditure at this rate, noting that the U.G.C. approved orders alone for 1964-65 was £1.2m. The following brief survey shows the consequences of a constant £2m. annual spend programme.
319. We have considered two extreme programmes proceeding on an annual hardware spend rate of £2m. There are two conflicting requirements to be met by our recommendations: on the one hand, to make computers available to as many institutions as possible; on the other, to increase the total computing power. We first investigate a programme emphasising numbers of computers rather than computing power by eliminating the large machines and stretching out the remaining programme so that it runs at the set figure of £2m per annum. Secondly, we keep all the large machines, but stretch the remaining programme even further.
This means that there would be no large machines installed in London, Cambridge, Manchester or Edinburgh. A survey shows that even if this were to be decided upon, many other Universities would have no new machines during the 5 years, and some would have to retain small machines indefinitely. A review shows that under these circumstances (for example) at least seven Universities would not have new computers during the five year period in question. Another five Universities would have to retain small computers which are admittedly too small for them at this stage, throughout the period.
The effect of concentrating on the large machines in London, Cambridge, Manchester and Edinburgh would be to reduce drastically the money available for new small and medium machines. A survey shows that at least fourteen Universities (for example) could not be provided for in any way during the next five years while five Universities would have to make do with small facilities for the next five years.
320. In practice one would presumably choose an intermediate course; but from what we have said it is now sufficiently clear that many Universities would have no computers, or very unsatisfactory ones, for the foreseeable future and the overall growth rate would be grossly reduced especially if any of the large machines are omitted. We repeat that we regard a programme continuing at a rate significantly lower than the one we have recommended would have extremely serious consequences both for the future of scientific research and of the computer industry in this country.
