ACD Atlas Computing Division ICF: Engineering Board Report

Jump To Main Content

Jump Over Banner

Home

Jump Over Left Menu

6. SOLUTIONS TO USER NEEDS AND RECOMMENDATIONS

Recommendations on solutions of future computing needs are made very difficult by the extraordinarily dynamic nature of the computer business. The power of mini computers is increasing rapidly and the cost for a given performance is falling dramatically. Progress in larger machines is less marked. A crucial question which is particularly relevant to interactive computing is whether power should be provided centrally or locally - or rather what proportion of each should be provided. Certain users require facilities that can only be provided locally - particular examples are where very high transmission rates are required between their terminals and the computer. Other users are so thinly spread that the only economic solution is to satisfy their requirements with a central facility. About 25% of the users considered in this study fall into each of these classes leaving 50% that can in principle be satisfied either locally or centrally.

6.1 Should Central or Local Facilities be Provided?

This discussion concerns about 50% of the user population that can in principle be satisfied by either central or local facilities.

6.1.1 Definition of Possible Solutions

Possible ways of providing interactive services are:

  1. Central service using a large machine serving 600 users connected by a star network who are provided with terminals in their university departments. About 60 input channels are provided allowing entry to many users at the same time.
  2. Local multi-user mini. This is a modest size machine serving 60 users with about 6 input channels and up to 6 simultaneous users.
  3. Local single-user mini. This is sometimes referred to as the personal computer. The machine serves one user at a time although it satisfies the requirements of a group of about 10 users.
  4. Cooperative computing. Cooperative computing allows a given application programme to be split between two machines and to run concurrently in both. This form of computing has not yet been successfully established and is therefore not considered further.

6.1.2 Arguments For and Against Central and Local Facilities

The arguments for local minis are:

  1. University groups prefer local facilities.
  2. It is possible to provide facilities requ1r1ng fast links if these become necessary.
  3. Scheduling, accounting and communications are easier.

The arguments against local minis are:

  1. Mini computers have a user address space limitation that is normally ≤64K. This is inadequate and rules out local minis at present.
  2. Satisfactory multi-user system software is not currently provided by manufacturers of minis. This is likely to change in the near future.
  3. It is more difficult to provide an integrated service and to reduce unnecessary duplication in software effort.
  4. There is limited processing power and it is not possible to satisfy the infrequent but sometimes important occasions when a user wants larger power.
  5. Because of queuing problems the average utilisation is reduced.
  6. It is expensive in staff costs to provide a routine service during dark hours on 10 minis on 10 separate sites. With a central machine it is economic to operate it during dark hours to provide a batch service and hence also to provide an interactive service for those few users that want exceptional conditions that cannot be met during peak hours.

6.1.3 Conclusions on Local versus Central Facilities

At the present moment central facilities are the only solution for the 50% user demand that could in principle be satisfied by local facilities. Thus in total 75% of the user demand should be met centrally and 25% locally for the immediate period - 1976-1978. Beyond this period it is probable that cheaper and better services will be provided by local minis. (See Appendix 2). With either solution a communication network will become more necessary with the passage of time.

Although minis cannot satisfy the requirements at this stage the two following sections discuss the capital and recurrent costs of central and mini systems at current prices.

6.1.4 Capital Costs

Capital has to be provided for the computer, terminals and communications. To make the comparison these costs have to be evaluated per user channel. Appendix 1 discusses communication costs. Conclusions are summarised in Table 6.1.

Terminal costs are the same in all cases but the costs are nevertheless included for completeness on the assumption of one alphanumeric terminal and one storage display system per channel at a total cost of £4K.

Table 6.1: Capital Costs per user channel for computers, terminals and communications.
Capital Costs in £K
Machine Communications
to Terminals
Terminals Total
Central System 16.7a 5.9d 4 26.6
Multi-user mini 13.3b 0.8d 4 18.1
Single-user mini 27.0c 2.3d 4 33.3

(a) £1M for 60 user channels - see Section 6.2.1

(b) £80K for 6 user channels

(c) £27K for one user channel

(d) See Appendix 1

6.1.5 Recurrent Costs

It would require 60 single-user minis to provide a service for 600 users. With 60 separate installations it is not possible to provide a similar service to that at a central facility and hence single-user minis are ignored below.

Items to be included in recurrent costs are:

The comparison to be made is one central system against 10 multi-user minis. The staff required to cover operations, applications, customer support and systems programming for the central system is about 20 (see Section 6.4.1). About the same total staff would be required for 10 multi-user minis.

Media costs would be the same per user for both systems. Maintenance costs are taken as 8% of capital costs per year. Housing costs are dependent on where the installations are put and what arrangement is agreed with the university.

The conclusion is that there are no major differences between the two solutions as far as recurrent costs are concerned.

Detailed recurrent costs for central facilities and multi-user minis corresponding to a recommended staffing are given in Section 6.4.

6.2 Hardware Recommendations for the Immediate Period (1976-78)

6.2.1 Recommendations for Interactive Computer Provision for Period 1976-78

It is recommended that:

  1. A central system of at least 3 × Atlas power should be installed at Chilton to satisfy the needs of 600 users. The system should have the following general specification.
    • Processor Power ≥ 3 × Atlas
    • 2 million bytes of main memory
    • Virtual memory capability and user address space of at least 1 Mbytes
    • 800 Mbytes of disc capacity
    • Capability of handling 60 simultaneous users
    A specification is given in Section 7. The estimated capital cost is £1M spread over two years. As the engineering research workers' need is for a reliable service, the requirement that the system should have been proven, including both software and hardware, in an engineering research multi access environment, should be given the greatest importance in selecting a machine. Major importance should not be attached to compatibility with existing equipment, maximum number crunching power or batch capabilities in making the machine choice.
  2. The UMIST PDP10 KA interactive system should be enhanced to provide a general service for 150 users. The facility should cease to be associated with a specific research project and should be a general SRC Engineering Board facility. The Director of the Rutherford Laboratory should be responsible for it. The existing machine should be expanded by adding:
    • 128 Kwords of memory
    • 4 × 10 Mwords of exchangeable disc capacity
    • A front end communications processor
    A specification is given in Section 7. The estimated capital cost is £160K.
  3. The Edinburgh PDP10 system should continue to be treated as a general SRC Engineering Board facility but should be expanded to provide a service for 250 users. The Director of the Rutherford Laboratory should be responsible for it. The facility should be enhanced by:
    • Replacing the existing KA10 processor by a KI10.
    • Adding 128 Kwords of memory
    • Adding 2 × 10 Mwords of exchangeable disc capacity
    A specification is given in Section 7. The estimated capital cost is £220K.
  4. Funds should be provided for the enhancement of 5 existing mini computers in university departments to bring them to multi-user capability. The aim should be for them to satisfy 50 users each. It is not possible to give a detailed specification of these enhancements until the actual machines to be enhanced have been selected. The staff at Chilton should assess the existing machines and make recommendations. The estimated cost is £160K.
  5. Two new multi-user minis should be purchased for development purposes. The Rutherford Laboratory should be responsible for this development in conjunction with appropriate university groups. The aim of this work should be to assess accurately the future capability of such systems rather than to provide facilities in the short term. A specification for these machines is given in Section 7. The estimated cost is £160K.
  6. Encouragement should be given to British manufacturers to provide suitable multi-user minis by the early notification of specifications and future requirements. This also applies to terminal and communication equipment.

Table 6.2 summarises the user service provided by these recommendations.

Table 6.2: Interactive facilities available by 1978
Location Computer CP Power
Atlas Units
Users
Supported
User Channels
Supported
Chilton ? 3 600 60
UMIST PDP KA10 0.7 150 15
Edinburgh PDP KI10 1.6 250 25
5 existing centres
to be enhanced
Various
(eg PDP11/45)
5×0.3 5×50 5×5
Total 6.8 1250 125

There are other computers that have been provided by SRC for interactive work but they are for special purposes or are sufficiently small to be ignored without distorting the overall picture. This plan provides 0.2 Atlas hours/week and 3.6 hours/week of connected time for each user. This falls well below the requirements given in Table 5.7 but is about a factor of three greater than is currently available. It is a reasonable first stage plan. It is possible that some of the shortfall may be satisfied by future Computer Board provisions. The effective processor power recommended for interactive use by 1978 is only 8% of the current batch processing power used by engineers on Computer Board machines. Further provision beyond 1978 is discussed in Section 6.5.

6.2.2 Recommendations for Provision of Terminals for Period 1976-78

Terminal requirements were discussed in Section 5.4.4. It is estimated that the cost is £4K per user channel. A possible breakdown of the provision is given in Table 6.3 but it is recommended that the expenditure be spread over the period 1976-78 and that the actual provision should be guided by experience of user needs. The total requirement for terminals to match the 125 channels in Table 6.2 is 250. Provision is made for 200 because of the existing terminals at UMIST and Edinburgh.

It is recommended that a sum of £540K be provided for terminals and hard copy units.

TABLE 6.3: Possible Terminal Provision to cover period 1976-78
Type of Terminal Example Cost/Unit
£K
Numbers to
be provided
Cost in £K
Alphanumeric terminals Teletype 0.9 100 90
Simple graphics terminals T4010 2.7 80 216
Larger and better quality graphics terminals GT40/T4014 6.5 20 130
Hard copy units for graphics terminals Tektronix hard-copy unit 2.2 20 44
Simple lineprinters Centronics 2.2 30 66
546

These terminals are for the central and local facilities. In particular some uses of the GT40 will require substantial local computer power.

6.2.3 Recommendations for Provision of Communication Equipment

The objective is to provide an integrated and coordinated interactive computing service to engineers. This is only really possible if all of the component parts are interconnected and this must be the long term aim. It will only be fully achieved if there is a unified network of computer links available to all universities. This provision is too expensive and ambitious for the SRC to undertake. It is recommended that the SRC together with other interested parties (Computer Board, DES, D of I etc) should encourage the Post Office to construct rapidly a national network. £50K of capital equipment and four staff at Chilton are recommended as a provision towards the network development involved in respect of the interactive facilities.

For the period 1976-78 a general network will not be available. It is recommended that the Chilton, UMIST and Edinburgh systems should be linked by 9.6 Kbits/sec private lines. The Chilton facility should be linked to the 360/195 and to the considerable SRC star networks already existing. As minis are upgraded to local multi-user facilities they too should be linked to the other facilities. Money has also to be provided for the communications between user terminals and the central facilities - these are estimated to cost about £5.9K per user channel. (See Appendix 1). Some equipment already exists but it is estimated that about 100 new channels will be required at a total cost of £590K. Including the links between Chilton, UMIST and Edinburgh the total recommended capital for communications is £620K.

6.3 Other Recommendations for Period 1976-78

6.3.1 Software

Development and documentation of application packages is a very costly process. At the moment there is considerable duplication in development work and comparatively little work is fully documented. The Technical Group recommends that a strong attempt be made to reduce the unnecessary duplication in development and to provide adequately documented software for general use. The provision of an integrated interactive service should help achieve this objective. Where possible existing software and packages should be used including industrial and commercial packages where relevant. The general policy for applications packages of limited use should be to provide packages at a single site and users can then gain access to these through the interconnection of all facilities. This single site philosophy should apply to special languages and special hardware as well as to application packages.

An adequate number of qualified staff should be provided at Chilton to assess user needs for specific application packages and to mount the selected packages at the relevant machine. The level of staff recommended in Section 6.4.1 is inadequate for any major production of user packages. Should a case be made and accepted for significant work in this area extra staff will need to be provided.

Preparation of some of the existing applications software for general service must be one of the first tasks undertaken. In particular an interactive FORTRAN library should be established. Co-ordinating and documenting for all the levels of graphics software should be started.

6.3.2 Bulletin

A Bulletin or Newsletter related to the work of the proposed interactive service should be established and published regularly. The Bulletin should have wide circulation to computer scientists in universities, polytechnics and other organisations, as well as to users and potential users. The contents of the Bulletin or Newsletter should be selected in order to make, amongst other things, a significant contribution to the identification of applications software development projects.

6.3.3 Meeting House Function

Personal contacts between the staff of the service, users, potential users, and computer scientists, should be encouraged. Resources should be made available to finance:

  1. Applications workshops.
  2. Regular conferences with publication of presented papers. Whenever possible, these activities ought to be held at different locations around the country.
  3. Development of the meeting house activity.

6.4 Staff Provision and Recurrent Costs for Period 1976-78

(All figures in this Section relate to direct staff)

6.4.1 Staff at Chilton

It is recommended that staff be provided at Chilton to operate the facility 24 hours per day and seven days per week. It is anticipated that the majority of the interactive load will be between 9.0 am and 10.0 pm on Mondays to Fridays and it is assumed that two operators will be required throughout this period. During the remaining period batch processing can be undertaken and interactive facilities can be provided for users requiring abnormal facilities such as large central processor power or large memory or file space. It is assumed that one operator will be required in these dark hours. Allowing for overlaps between shifts, holidays etc a total operating staff of 9 will be required together with a service manager.

It is recommended that a group be established to provide user support. This group will be responsible for documentation, answering user enquiries, identification and education of new user groups, routine provision of established packages, the newsletter/bulletin, organisation of workshops and meeting house activities. A staff of 6 is recommended for this group.

A staff of 3 is recommended for systems programming and 6 for communications and network development.

In Section 6.2.1 it was recommended that two new multi-user minis be provided to enable development work to be undertaken by the centre to assess the future capabilities of such systems. A staff of 4 is recommended for this work. This group should also give advice and assistance to universities provided with multi-user minis.

In Section 6 it was recommended that staff be provided to assess user needs for application packages and to mount the selected packages on the relevant machine. This function is logically combined with a general one of advising the Engineering Board, its committees and the proposed Engineering Computing Facilities Committee on various matters. Examples of this are: provision of information for the audit of management and usage of facilities provided through the central service, advice on grant applications involving computers and computing, identification of appropriate minis for expansion to multi-user systems. These functions will require staff of broad knowledge and experience. The staff will need to spend a great deal of time at user sites. A staff of 6 is proposed.

It is assumed that the whole of the responsibility for the provision, implementation and operation of the central facility and the overall responsibility for those facilities provided at UMIST, Edinburgh and elsewhere in the universities will be the task of a senior member of the Rutherford Laboratory. As such he will be required to participate in the management of the Rutherford Laboratory and to represent there the views of the user community.

The service offered by the staff at Chilton to engineers in universities will present difficulties from several points of view. First, the facilities needed by different areas of research will be different and staff at the Centre will need to understand these differences. Secondly, the needs of the research workers may change from time to time and it will be important that the service offered is sufficiently flexible. Thirdly, it may be impossible in some circumstances to supply exactly what is desired by the research worker and effort will be needed to produce the nearest approximation to his needs that is possible. All of these difficulties will require a good deal of tact if friction is to be avoided. The management of the interactive service therefore needs to be in the hands of a man who has some experience in engineering research, preferably in a university environment. He must have a wide knowledge of interactive computing, extending if possible to interactive graphics. He must be able to deal with difficult and conflicting requirements requested by users. To find such a man will not be easy, but an unsuitable appointment for this position could easily jeopardise the success of the service.

The total direct staff recommended at Chilton is 36. It is assumed that the staff build up will start in 1975 and will reach the proposed level in 1978. The assumed build up is given in Section 6.4.3. It is assumed that it will remain constant at 36 beyond 1978 (See Section 6.5).

6.4.2 Staff in the Universities

It is assumed that the UMIST and Edinburgh systems will be operated from 9.0 am until 10.0 pm on Mondays to Fridays. A single operator is assumed resulting in a staff of 3 for operations plus a service manager. User support staff of 2, a systems programmer and a person to deal with communications are also recommended. This is a total staff of 8 at each centre which is about the current level provided through the existing grants to the Universities. Recommendations on the management of the facilities and the possible mechanisms for the support of them are discussed in Section 8.

It is difficult to assess the staff requirements for the 5 local multi-user minis until the systems to be enhanced are identified. For planning purposes an additional staff of 5 is shown.

6.4.3 Summary of Staff Provisions

A summary of the staff recommendations from Sections 6.4.1 and 6.4.2 are given in Table 6.4.

TABLE 6.4: Recommended Direct Staff for 1978
Function Chilton UMIST Edinburgh 5 multi-user
minis
Service manager 1 1 1 5
Operations 9 3 3
User support 6 2 2
Systems programmers 3 1 1
Communications and
networks development
6 1 1
Application advice to
Boards and Committees
and audit function
6
Multi-user minis
development
4
Division Head 1
TOTAL 36 8 8 5

The assumed build up of direct staff for the period 1976-78 is given in Table 6.5. Projections for beyond 1978 are given in Section 6.5.

TABLE 6.5: Recommended Provision of Direct Staff for 1976-78
Site 1976
man years
1977
man years
1978
man years
Chilton 12 30 36
UMIST 8 8 8
Edinburgh 8 8 8
Local minis 1 2 5
TOTAL 29 48 57

6.4.4 Recurrent Costs

Provision has to be made for the following items in estimating recurrent costs:

  1. Direct staff costs (evaluated at £5.3K per man year).
  2. Overheads including indirect staff, electricity, heating, and all general services (evaluated as 100% of direct staff costs).
  3. Maintenance of computers, terminals and communication equipment (evaluated at 8% of capital cost/year).
  4. Rental charges including Post Office private lines, provision of software, purchase of time on computers where special facilities can be provided (an example of this could be time on the NEL 1108 to provide access to special packages such as NASTRAN and APT). It is very difficult to provide estimates for these costs until the actual user community is determined. For budgetary purposes it is assumed that the Post Office rentals in 1978 will be £60K (corresponding to 30 private lines of an average length of 100 miles). Other rental items are assumed to be £50K in 1978.

6.4.5 Summary of Capital and Recurrent Costs for 1976-78

A summary of the recommended capital expenditure for the period 1976-78 together with the staff and recurrent costs proposed is given in Table 6.6.

TABLE 6.6: Estimates of Capital and Recurrent Expenditure for 1976-78
1976 1977 1978
Capital
£K
Recurrent
£K
Capital
£K
Recurrent
£K
Capital
£K
Recurrent
£K
CHILTON
Computer hardware 700 300
Mimi time-sharing systems 80 80
Network development 15 15 20
Direct staff 64 159 191
Overheads 64 159 191
Maintenance 8 71
TOTAL 95 128 795 326 320 453
UMIST
Computer hardware 160
Direct staff 42 42 42
Overheads 42 42 42
Maintenance 30 30 43
TOTAL 0 114 160 114 0 127
EDINBURGH
Computer hardware 220
Direct staff 42 42 42
Overheads 42 42 42
Maintenance 30 30 48
TOTAL 0 114 220 114 0 132
UPGRADE OF EXISTING SRC MINIS
Computer hardware 40 60 60
Direct staff 5 11 27
Overheads 5 11 27
Maintenance 3 8
TOTAL 40 10 60 25 60 62
GENERAL
terminals 120 220 200
Terminal Maintenance 0 10 27
Communications hardware 150 250 220
Communications maintenance 0 12 32
PO line rentals 10 60 60
Other (software rental etc) 5 25 50
TOTAL 270 15 470 107 420 169
GRAND TOTAL 405 381 1705 686 800 943

TOTAL CAPITAL IN ROUND TERMS = £3.0M

TOTAL RECURRENT IN ROUND TERMS = £2.0M

GRAND TOTAL = £5.0M

6.5 Recommendations for the Period beyond 1978

6.5.1 Shortfall in Hardware Provision

The hardware provisions recommended in Section 6.2 (see Table 6.2) fall a long way behind the user needs identified in Section 5.4 (See Table 5.7). Table 6.7 compares the need and the recommended provision and identifies the shortfall.

TABLE 6.7: Comparison of User Needs and Hardware Provisions including those recommended for the period 1976-78
1975 1976 1977 1978 1979 1980
Processor power needed (Atlas units) 7 10 15 20 24 28
Channels needed 126 144 162 180 180 180
Terminals needed 252 288 324 360 360 360
Processor power available (Atlas units) 2 2 5 7 7 7
Channels available 30 50 90 125 125 125
Terminals available 50 90 170 250 250 250
Processor power shortfall (Atlas units) 5 8 10 13 17 21
Channel shortfall 96 94 72 55 55 55
Terminal shortfall 202 198 154 110 110 110

It is clear that considerable further provision is necessary particularly in processor power. It is difficult to make specific recommendations on how to meet the needs for the following reasons:

  1. The rapid improvement in the processor power and cost of mini computers together with anticipated improvements in multi-user software, user addressable space and decrease in the cost of memory make it very probable, but not certain, that within two years the multi-user minis will be the best way to enhance the interactive computing facilities (See Appendix 2).
  2. In the longer term it is anticipated that improvements may even have reached a stage where the single user mini is a viable alternative. For this to be so a general computer network must be available to allow all users access to all facilities. Without this the great advantages of an integrated system will be lost.

6.5.2 Recommendation for Period Beyond 1978

The Technical Group strongly recommends that a further review be made in 1977/78 to assess what future provisions be made. By then

  1. Experience will have been gained with the central service.
  2. A clearer picture will have been established of the real user needs.
  3. The enhancement of local minis to multi-user minis and the development work with the two minis provided will give a great deal of information about multi-user minis.
  4. The experimental programme on networks including the considerable SRC effort will have removed the current uncertainties and it is hoped that the Post Office will have defined what is to be available, the time scale and the charges.
  5. If the current trends have continued the cost of minis will have decreased and better systems will be available.

There is clearly a need to provide a budget estimate of what provision should be made for capital and recurrent expenditure for the period 1979-81. To do this it is necessary to assume a model solution. The assumption is made that multi-user minis will be provided from 1977-81. An overlap is taken with the period already considered (1976-78) because it is expected that the two minis provided for development can be installed in selected universities in 1977/78 and additional systems can be bought from 1978 onwards as the picture develops and the case is made. It is assumed that averaged over the period 1978-81 a multi-user mini of 1 Atlas power together with necessary file store and terminals will cost £50K. (At current prices it would cost about £lOOK). If this assumption is true and provision is made for 20 units the capital requirement for computers and terminals is £lM. Additional capital will be required for network hardware and this is assumed to be £150K over the 3 years. The communication equipment already purchased will still be required for connection into the network.

The recurrent costs are even more difficult to estimate but assuming the staff at the centre remains fixed at 36 and the staff in the universities increases by 20 over the period the recurrent costs are estimated to reach £1.3M per year by 1981.

Table 6.8 provides a breakdown of these additional estimates between 1978 and 1981 and Table 6.9 combines the costs with those given in Table 6.6. The total capital cost is £4.0M and the recurrent totals £5.6M giving a grand total of £9.6M for the period 1976-81.

Table 6.8: Budgetary Estimates of Additional Provisions for Period 1978-81
1978 1979 1980 1981
Capital
£K
Recurrent
£K
Capital
£K
Recurrent
£K
Capital
£K
Recurrent
£K
Capital
£K
Recurrent
£K
Provision of further multi-user minis and terminals 100 300 300 300
Capital for networks and communications 50 50 50
Staff costs at Chilton
included in
Table 6.6
for period
1976-1978
191 191 191
Staff costs at universities 143 175 212
Overheads at Chilton 191 191 191
Overheads at Universities 143 175 212
Maintenance 299 327 355
Rentals and line charges, including network charges 120 130 140
TOTAL 100 350 1087 350 1189 350 1301
Table 6.9: Estimates of Capital and Recurrent Expenditure for 1976-81
1976 1977 1978 1979 1980 1981
Capital
£K
Recurrent
£K
Capital
£K
Recurrent
£K
Capital
£K
Recurrent
£K
Capital
£K
Recurrent
£K
Capital
£K
Recurrent
£K
Capital
£K
Recurrent
£K
Computers and terminals 240 1440 660 300 300 300
Communications and networks 165 265 240 50 50 50
Staff costs at Chilton 64 159 191 191 191 191
Staff costs at universities 89 95 111 143 175 212
Overheads at Chilton 64 159 191 191 191 191
Overheads at universities 89 95 111 143 175 212
Maintenance 60 93 229 299 327 355
Rentals and line charges including network charges 15 85 110 120 130 140
TOTAL 405 381 1705 686 900 943 350* 1087 350* 1189 350* 1301

* The low capital included in this period is dependent on price reductions of a factor of 2 by 1980 for mini computers.

Total Capital in Round Terms = £4.0M

Total Recurrent in Round Terms = £5.6M

GRAND TOTAL = £9. 6M