The following brief comments give the individual views of a number of designers and research workers, with experience of interactive design, on the scope for further research in their own areas of interest. The subject areas included are not intended to be comprehensive.
Because of the essentially iterative nature of design virtually all research being carried out into computer-aided architectural design (CAAD) has been, and is likely to continue to be, based on the development and use of interactive systems. Perhaps more than any other design activity, architectural design is governed by a multitude of legal and quasi-legal regulations. Frequently changing and difficult to memorise, these regulations are best incorporated in computer programs which can be recalled at relevant points in the design process and it is likely that future research into CAAD methods will be aimed at discovering ways of speedily presenting designers with the consequences of the interaction between their own decisions and the effects of the regulations.
As the traditional and firmly entrenched means of communication in architecture and building is by means of drawings, any advanced CAAD system will be based on drawing and sketching and it is certain that, for some time to come, a great deal of detailed research will be needed in the design of responsive systems conducive to the architect's graphical method of working.
The importance of experience and intuition in engineering practice is well recognised and in Civil Engineering it is probably more significant than in any other branch of engineering. It is impossible or impracticable to represent formally the logic of all decision making such that it can become an automatic part of a computer program or computer system. This is true even with large modern computers and it is relevant to decision making at the level of detail design or in the overall control of construction projects.
The availability of interactive computation facilities enables provision to be made of a working environment in which the engineer and the computer fulfill the functions which they can each do most efficiently. Much of the information concerned can with considerable advantage be represented graphically in the form of curves, drawings or pictures. The interactive use of visual displays should therefore be an important and an integral part of engineering computation.
The small amount of work that has already been done in developing interactive computer systems in Civil Engineering has shown them to be economically attractive and to provide considerable advantages in time, an important factor in engineering construction. There is a great need therefore to encourage work in this area and it is clear that if interactive computation is to achieve its full potential a great deal of this work must be done in the universities. There will also be a valuable spin off in the use of the software for teaching purposes, particularly in engineering design, by enabling students to see rapidly the effects of changing parameters in a variety of design situations.
One useful application lies in the area of structural steelwork design and detailing: interactive computer graphics enables the solution of complex geometrical detailing problems without recourse to the preparation of full scale detail drawings as hitherto. This application has shown the importance of having a compatible batch and interactive system so that straightforward situations can be recognised and dealt with automatically.
Some of the first applications of interactive computation have been in the field of structural engineering. There are however many other useful applications in the broader field of Civil Engineering, the most important being in the interpretation of satellite and aerial photographs involved in the control and management of water and mineral resources and also in the acquisition and use of data for urban and regional planning including route determination. There is also an extremely large number of design applications including problems in foundation and pavement design, water supply and sewage disposal, traffic engineering and vertical and horizontal road curve design, in addition to structural engineering.
Communication with the computer has been one of the greatest difficulties in computer development over the last twenty years and interactive computation provides the most promising step that has yet been made towards ease of communication.
Control systems, even relatively small ones, can have very complicated behaviour. Their performance also has to satisfy a large number of requirements; some of these can be formulated mathematically (eg stability) but others are qualitative (easy availability of components, compatibility with equipment already installed, etc). Many methods proposed in the current literature, such as optimal control, or pole assignment, ask the designer to specify what he wants in such detail that there is only one solution. This solution is then found by an algorithm in the computer.
Such an approach has several disadvantages:
These considerations suggest that design methods need to pay attention to the efficient use (in the widest sense) of the designer, as well as to the efficient use of the computer. Some parts of the design procedure are best carried out in the designer's mind (pattern recognition, innovation, etc) while others are more suitably done in the computer (algorithmic computations, logical checking, etc). Such a division becomes possible when easy communication is provided between the designer and the computer.
These conclusions, which are not unique to control, give strong support for the suggestion that industrial applications of control will need CAD techniques. Since little work has been done so far in developing these, university research in this area will be important for many years.
The finite element method was pioneered in the field of structural analysis where it is now used extensively. Over the last few years increasing interest has been shown in the analysis of other field problems such as magnets (Rutherford Laboratory) and environmental pollution (Southampton University). For about the first 15 years of the method's history, emphasis in research centred on equation-solving techniques, convergence criteria and refining element representations. This kind of research continues, especially in these new areas of application. For the last 5 years it has been increasingly recognised in industry that the current need is for improved techniques for input and output. Data preparation has in many cases become more expensive than the computer costs of analysis, and stress men have for years been drowning in paper output. There are then three areas in which man/machine interaction promises to yield benefits.
Evidence of trends in computer hardware costs and manpower costs demonstrate the increasing benefit of the interactive approach in these areas. However, the field of application is very broad indeed and extensive interactive engineering computing research and training will be needed to yield, in the first instance, aids to current procedures before more comprehensive computer aided design facilities can be established.
At present three distinct needs can be identified.
Computers are widely used to construct numerical approximations describing particular fluid flows. The sizes of problems considered range from small ones that can be done with desk calculating machines to large problems which cannot yet be solved accurately even with the aid of the most powerful existing computers. The determination of flow behind a given shock wave in supersonic flow is an example of a small problem involving only the solving of simple algebraic equations. Turbulent flows, described by a system of four or more non-linear partial differential equations having time and three-space co-ordinates as independent variables, are typical of flows which cannot yet be calculated satisfactorily. Between such extreme cases there are, of course, many fluid flow problems that can be solved with the aid of existing computers. Certain aircraft wing-design problems and the design of wall shapes for diffusers and contractions in ducting are typical of this class of problems.
In the most general fluid dynamics problems of interest in engineering, geometrical structures of three-dimensional flow fields which occur may be complicated, depending on the shapes of boundary surfaces.
Interactive computing facilities are needed, and will be used, to generate suitable forms of input data sets describing starting solutions or the shapes of boundaries containing flow fields; to obtain visual displays of flow patterns generated numerically by computers; to obtain visual displays of numerical representations of boundary surfaces and associated co-ordinate surfaces in space; to aid research concerned with new applications programs; and to carry out research aimed at developing more economical computing procedures than those used at present for solving large scale fluid dynamics problems numerically.
Access to the largest available number-crunching computers via links from an interactive computing facility, and guarantees of rapid job turnround on such number-crunchers, will make it possible to engage in applications software research on a partially interactive basis. For example, by changing parameters at intermediate stages of a calculation, computer users will be able to interact with calculations in a manner which hitherto has not been possible. Significant savings in the computer time used, and savings in the computer users' time, can be achieved. The ability to work on a partially interactive basis is a significant practical development which will improve our capacity to do useful research concerned with applications of fluid dynamics in many branches of engineering.
Computers have been used extensively in internal combustion engines and turbo machinery applications by designers in off-line calculations which are of closed form, that is, where a given configuration is entered into the calculation and the performance, or strength, or flow conditions etc are computed. Where there might be a range of possible configurations for a given design of component, there will obviously be a range of resultant answers to the problem and the computer turnround and cost may become excessive. To use the computer economically under these conditions, the designer's selection is clearly limited. If the calculations can be opened up so that the designer is interacting with the calculation, he can use his own initiative so that he might in fact evolve the design as the calculation proceeds. This would not only be cheaper but would also enable a better design to be obtained. Furthermore, the preparation of such a program would be simpler, since with the closed form of calculation the programmer has to attempt to cover all possibilities within the calculation if computational failures are to be avoided.
In internal combustion engine and turbo machinery design the range of computer usage is quite wide. In the first stage of the design one is normally concerned with the arrangement of components, performance predictions, overall dimensions, weight and cost. For a selected given engine configuration, detail consideration is then given to the component parts leading to fluid flow studies, heat flow studies, dynamic and static strength calculations, and control system arrangements, to meet the performance requirements. Whilst it is current practice for the computer to be used mainly for calculation purposes from previously prepared design sketches and drawings, recent advances in computer aided design have integrated the drawing and calculation phases and in some cases production, by producing tape or card outputs in a form for direct utilisation in numerically-controlled machine tools. This is an area where major advances can be expected. Another area which could reduce costs in engine design is the development of computer aided arrangement drawings for integrated systems with associated parts list, ordering and production schedules, cost estimates and weights. Preliminary work in this field is being carried out for ship pipe layout.
The ship design process is characterised by a high output of data and information, largely of a graphics nature, from relatively small design teams.
Already, computer aided design methods are extensively used in shipbuilding for initial design calculations, for extensive stress analyses, and for lofting and preparing production instructions for numerically-controlled machine tools. In all these areas, interactive working has proved to be very beneficial and is already in common use.
New and impending developments in analysis and in design concepts in many areas, both for the complex steel structures and for the various engineering systems, are beginning to put an increased burden on the design teams, which can probably best be met by extending the same approach.
Ship structures are complex three-dimensional assemblages, which are not only important in themselves, but interact with most other aspects of ship design. The generation and visualisation of these structures is a major problem and there is a need for the development of cost-effective processes in the following potential areas for interactive CAD:
For the engineering systems the main problems requiring an interactive CAD approach are concerned with:
Electronic design is an area in which interactive computing is particularly appropriate. The design process has many distinct phases from the initial concept to its physical realisation. Most of the relevant analytical steps and some parts of design synthesis can be catered for by automatic programs, but there remain substantial tasks for the human designer especially in initial and detailed synthesis. Simulation is important and may reflect various levels of complexity, eg a probabilistic system model, a digital logic model or a circuit model based on the physical characteristics of devices and their connections.
In manufacture, planar technologies are increasingly dominant. For example, printed-circuit boards consist of a number of connection layers each of which must be planer, and in integrated circuits each layer is of great geometrical complexity. Attempts to develop fully automatic programs for detailed layout have been disappointing and many electronics companies rely on interactive design with minicomputers and displays. It is important to develop more powerful automatic aids for use with interactive computing and this is a promising research area.
The design of electronic components and systems is important to research workers in other fields who may require, for instance, special-purpose digital systems for experiment monitoring and control, or in special circumstances custom-made integrated circuits may be called for. There is considerable scope for developing interactive computing systems to meet the general needs of research workers in this area.
Apart from very simple arithmetical calculations, the tradition in the textile and fibre industries is a qualitative attack on problems (on the whole, highly successful). For many reasons one can now see a need though scarcely yet much of a demand - for more numerical calculation. Interactive computing is the key to the problem.
Interactive computing in Textile Technology is needed for the following purposes:
My research experience and contact with the industry leaves me strongly of the view that if the necessary framework can be established by academic research, the methods will be taken up and prove of great benefit to industry.
The design and manufacture of engineering equipment is the industry that has fed and clothed us as a nation for more than a century, and it must be clear to most people by now that we have ceased to be as good at it as we once were, and that without some kind of renaissance it will not be doing so for much longer. British engineers still have a reputation for the excellence of their engineering concepts, but not infrequently the manifestation of these as hardware is marred by lateness, lack of attention to detail and consequent unreliability. This is by no means universal, but it is a reputation that sticks and does not improve our image vis-a-vis foreign competition.
British companies can rarely afford to spend the man-hours on a new design to which their counterparts in, for example, the United States are accustomed. Lower volumes simply do not allow this, yet the products are expected to be of comparable quality of detail design and reliability. To survive, we need to be just that much cleverer and extract more nearly a quart out of a pint pot.
To do this effectively requires quite different methods in design and, more particularly, in the transfer of the design into and through the manufacture. This is where interactive computing offers possibilities which, if grasped early, could have a catalytic effect. What are needed, are tools which enable an engineer to spend his time more effectively: in the concept stage to examine more exhaustively the basic trade-offs and evaluate a chosen compromise more thoroughly than would be possible at present; to specify the detail more swiftly and with fewer man-hours than is at present practical, and having done this, to have sufficient information stored ready in the pipeline to enable the manufacture of prototypes to begin and, after subsequent verification and change as necessary, to facilitate full-scale manufacture with a minimum of tooling. Changes which are inherent in any design/manufacture programme could be made easily without the long tedious error-creating chain which we have at present.
Many of the elements of this process already exist: they have been developed in the field of interactive graphics (visual concepts are the basis of all engineering design and much of the transfer of information) and in the fields of automatic drafting and numerically-controlled manufacture, which can itself be extended well beyond its present frontiers. Much of the hardware exists but considerable effort is needed to develop software in a number of realistic environments to create systems which will enable a worthwhile fraction of the engineering industry to achieve better ways of making things.
The computer will never be a substitute for the knowledge and skill of the designer, but it can be used to enhance these skills, to speed them up and to make them more thorough, all at the same time. At subsequent stages it can diminish routine drudgery and provide swifter and more accurate information transfer than the present tortuous chain.
Research in Artificial Intelligence is devising computer algorithms for performing tasks which people find intellectually demanding, like proving mathematical theorems, playing chess or writing programs, also for performing tasks which come easily to us but really require much sophistication, like conversing in natural languages or analysing everyday scenes. The algorithms are complex and have to embody knowledge in the form of rules or procedures to cope with a large number of particular cases, for example special constructions of English grammar, so that their development needs all the help that a good interactive time-sharing system can give. The programs produced are in a sense theories of intelligent behaviour, and they have to be created in the hypothesise-and-test manner typical of scientific theory making, their behaviour being scrutinised at each stage. Good man/machine interaction speeds up this incremental development process by such a large factor that almost all serious work in Artificial Intelligence has been done on interactive time-sharing systems.
As well as these general considerations some application areas of Artificial Intelligence have an experimental nature which requires on-line connection of apparatus and visual display facilities. These applications include analysis of scenes using TV cameras, speech recognition, computer-controlled assembly of parts and experiments in computer aided education for children.
A questionnaire was circulated to 372 engineering departments of universities and polytechnics to determine the number of staff and students engaged in engineering research. Departments were also asked to estimate numbers of staff and students working on SRC supported work, now and in 1978. Replies were received from 203 departments and Table 5.1 gives the results of the enquiry.
| Current number of professional staff (that is academic staff, professional staff, research fellows and research assistants) | 5254 | 8930 |
| Current number of research students | 3676 | |
| Current number of professional staff working on SRC funded work | 974 | 2120 |
| Current number of students working on SRC funded work | 1146 | |
| Estimated number of professional staff working on SRC funded work in 1978 | 1302 | 2925 |
| Estimated number of students working on SRC funded work | 1623 |
No allowance has been made in Table 5.1 for the 169 Departments that have not replied and hence it should be taken as a lower limit on the population.
The estimated number of staff in 1978 has not been taken to mean a growth in the number of posts that should be supported but rather a growth in the proportion of staff that will need access to interactive computing facilities.
The original Working Party Report identified three broad categorisations of computer usage by engineers. These were:
Interactive computing is crucial to category b and the Technical Group assumed that 80% of engineers involved in design methods research should have access to interactive computing.
Users in category a (large-scale analysis) accept that their main programs cannot reasonably be run interactively, but find interactive computing of great value for data preparation and for the analysis of results. The Technical Group again assumed that 80% of engineers involved in work in this category should have access to interactive computing, but weighted this requirement by a factor ½ to take account of the smaller interactive load.
Computers for data acquisition were considered outside the scope covered by this report but some engineers engaged in this area require interactive facilities in their data reduction and analysis. The Technical Group assumed that 20% of engineers in this area should have access to interactive computing.
An examination was made of the use of the SRC and Computer Board major computers by engineers. Machines of the ICL 1906A power and above were considered and the figures normalised to hours of IBM 360/195 time per week. Table 5.2 gives the results of the survey.
There are many smaller machines in universities which have not been included and extrapolation suggests that the engineers' use of Computer Board machines is equivalent to about 100 hours of 360/195 time.
The accuracy of the figures is limited by many factors but still indicates that the current use of batch processing by engineers approaches full use of a 360/195.
| Institution | Computer | Equivalent 360/195 hours per week |
|---|---|---|
| SRC COMPUTERS | ||
| Atlas Computer Laboratory | ICL 1906A | 1.0 |
| Rutherford Laboratory | IBM 360/195 | 1.3 |
| COMPUTER BOARD MACHINES | ||
| Imperial College of Science and Technology | CDC 6400 | 7.7 |
| South West Universities Computer Network | ICL 4/75 | 1.4 |
| University of Belfast | ICL 1906S | 1.5 |
| University of Birmingham | ICL 1906A | 3.6 |
| University of Cambridge | IBM 370/165 | 1.3 |
| University of Edinburgh Regional Computer Centre | ICL 4/75(2) | 0.4 |
| University of Edinburgh Regional Computer Centre | IBM 360/158 | 1.2 |
| University of Leeds | ICL 1906A | 1.3 |
| University of London Computer centre | CDC 7600 | 39.4 |
| University of London Computer Centre | CDC 6600 and CDC 6400 | 2.7 |
| University of Manchester Regional Computer Centre | CDC 7600 | 29.0 |
| University of Newcastle | IBM 360/67 | 1.7 |
| University of Nottingham | ICL 1906A | 5.0 |
| University of Oxford | ICL 1906A | 0.4 |
| TOTAL USAGE | 98.9 | |
1 hour on IBM 360/195 is equivalent to 30 hours on Atlas
The University of Leicester CDC Cyber 72 and University of Liverpool ICL 1906S were not included as they had not been in use for long.
The comparison of the machines is based on the speed of execution of optimised FORTRAN code. No account has been taken of any other factors such as operating systems and languages used.
Nine computer installations costing over £50K have been provided by the Engineering Board to university departments for interactive work. These are listed in Table 5.3, discounting a hybrid machine at Imperial College.
| Institution | Computer |
|---|---|
| Imperial College of Science and Technology, Department of Chemical Engineering | Honeywell DDP516 |
| Imperial College of Science and Technology, Computing and Control Engineering Department | DEC PDP15/40 |
| University of Cambridge | CTL Modular One |
| University of Edinburgh | DEC PDP KA10 |
| University of Manchester, Institute of Science and Technology | DEC PDP KA10 |
| University of Oxford, Computer Science Department | CTL Modular One |
| University of Southampton, Institute of Sound and Vibration Research | DEC PDP 11/70 |
| University of Warwick, Inter-University Institute for Engineering Control | RXDS Sigma 5 |
| University of Leicester, Department of Engineering | DEC PDP 11/45 |
The machines have a wide variety of peripherals, terminals and experimental equipment. Apart from the two PDP10s there was little common software owing to the variety of machines and uses. A total of about 300 users are supported on these machines.
Table 5.1 in Section 5.2 shows that a lower limit of engineers engaged on SRC supported work is currently approximately 2100 and it is estimated that it will rise to approximately 3000 by 1978.
Information gathered during the user visits has been used by the Technical Group to estimate what fraction of engineers would be engaged in the three categories of work described in Section 5.3.1 if facilities were available. The conclusions are given in Table 5.4 for now and 1978.
| Fraction of Engineers |
Engineers working on SRC funded work |
Percentage needing interactive facilities |
Potential users |
||||
|---|---|---|---|---|---|---|---|
| 1975 | 1978 | 1975 | 1978 | 1975 | 1978 | ||
| ENGINEERS USING COMPUTERS | |||||||
| Large-scale analysis | 1/6 | 350 | 500 | 40 | 140 | 200 | |
| Design method research | 1/3 | 700 | 1000 | 80 | 560 | 800 | |
| On-line experimental work | 1/3 | 700 | 1000 | 30 | 210 | 300 | |
| ENGINEERS NOT USING COMPUTERS | 1/6 | 350 | 500 | 0 | 0 | 0 | |
| TOTAL | 1 | 2100 | 3000 | 910 | 1300 | ||
The values given in Table 5.4 are meant to indicate the scale of demand rather than to represent a precise estimate. Computer Board facilities do not meet this demand nor can they meet the project-oriented requirements of engineers in this area.
Experience at UMIST, Edinburgh and other universities reasonably supplied with interactive computing facilities indicates that each user currently requires on average 0.25 hours of an Atlas power machine per week. This experience is also confirmed by commercial organisations. Thus in total the current demand for effective computing power on interactive computers is of the order of 230 Atlas hours/week.
The nature of interactive computing is incompatible with very high central processor utilisation if a satisfactory response is to be provided - a figure of 60% is reasonable. Moreover only about 60 hours per week are effective for interactive use. Thus a 1 Atlas power central processor gives 36 Atlas hours of interactive computing per week. The remaining time can be used for batch processing.
Extrapolations for the future are g1ven in Section 5.4.5.
A further relevant statistic from existing installations is that the average CP power required per logged in user is about 0.035 of an Atlas, but the distribution is very skewed with a small number of users requiring much larger powers. Some users working in large scale analysis require access from interactive facilities to the largest available computing power.
From an examination of sample engineering users programmes it appears that the proportion of programmes (including data area) which will fit into a given memory space with and without overlay or chaining is that given in Table 5.5.
| Memory Kbytes |
Percentage fitting in without overlay |
Percentage fitting in with overlay |
|---|---|---|
| <25 | 10 | 20 |
| <60 | 30 | 70 |
| <150 | 60 | 90 |
| <400 | 90 | 95 |
| <1000 | 98 | 98 |
However, overlay/chaining slows down the progress of the engineering research and demands the inclusion of skilled programming effort in research teams thus adding substantially to the cost. A better solution is to provide machines with virtual memory having sufficient real memory to allow average user programs of the order of 150 Kbytes. Users should have access to at least one machine with a user-addressable area of 1 Mbytes.
Experience on available installations indicates that a file store of about 1 Mbytes of on-line storage should be provided per user. In addition about 100 Mbytes is required for program libraries and system use.
Little evidence was found of the need for large databanks (capacity 100-1000 Mbytes) in university engineering research although these are required in industry by those working in areas such as the design of large integrated systems, purchasing, etc. In the university area the largest requirement is in architecture where there are demands for data files running into tens of Mbytes.
Current experience on interactive systems indicates that each user requires about 5 hours of connected time/week and that one channel port into the system can on average support 10 users. This requires the system to be available at least 80 hours/week as it is not possible to achieve much more than 60% utilisation of channels on average. Small systems can support less users/channel because of queue problems.
There are a wide range of possible terminals and Table 5.6 lists some of them with relevant properties.
| Type of terminal |
Examples | Approximate cost in £K |
Required line speed Kbits/sec |
Typical users |
|---|---|---|---|---|
| 1 Alphanumeric | Teletype, typewriter, visual display unit | 0.8 | 0.1 - 0.6 | Programme development Non-graphical interaction |
| 2 Storage display tube and keyboard |
Tektronix 4010 or large screen T4014 | 2.5 (4010) 6 (4014) |
1.2 - 9.6 | Graphs and diagrams moderate interaction rate |
| 3 Simple refresh systems with local store |
GT40 | 9 | 1.2 - 9.6 | Graphs and diagrams Faster interaction, grey scales |
| 4 Refresh systems for dynamic displays including lightpens or equivalents |
Vector General or Evans and Sutherland | >20 | 50 | Full dynamic displays Fully interactive |
(1) The recommendations in this report are based upon a service normally limited to a transmission rate of 4.8 Kbits/sec.
(2) The above table is a considerable simplification of a complicated subject.
In addition there are various hard copy devices including line printers, hard copy units for storage displays, high quality film and microfilm recorders.
Terminals and hard copy devices connected to computers over distances exceeding about half a mile are for practical and financial reasons limited at present to line speeds of 9.6 Kbits/sec
Communication costs increase rapidly as the transmission speed is increased. User needs exist for all forms of terminals but the majority of users can be satisfied with line speeds ≤ 4.8 Kbits/sec. Those requiring the highest speeds will have to use terminals directly connected to computer installations either by provision of local systems or by going to a centre where such facilities are available.
Experience indicates that about 2 terminals are required per input channel (ie one terminal per 5 users) and that about 50% should be character terminals and 50% storage display tubes or equivalent simple refresh systems with local store. Hardcopy units need to be provided at a level of about one per 4 other terminals although this is dependent upon the actual grouping of users and their particular application.
Table 5.7 summarises the numbers of engineers working on SRC funded research work requiring interactive facilities, and their needs. The number of users has been evaluated in Section 5.4.1. Evidence available suggests that the processor power required per user will grow from the present 0.25 Atlas hours/week to 0.75 by 1980. It is assumed that the requirements for connected time will remain at 5 hours/week. The justification for the increase in use of interactive computing is that the present demand reflects the shortage of facilities and is less than the usage by engineers in industry. A deliberate attempt must be made to correct the balance.
| 1975 | 1976 | 1977 | 1978 | 1979 | 1980 | |
|---|---|---|---|---|---|---|
| Number of interactive users | 910 | 1040 | 1170 | 1300 | 1300 | 1300 |
| CP power/user in Atlas hours/week | 0.25 | 0.35 | 0.45 | 0.55 | 0.65 | 0.75 |
| Required CP power in Atlas units | 7 | 10 | 15 | 20 | 24 | 28 |
| Connected time/user hours/week | 5.0 | 5.0 | 5.0 | 5.0 | 5.0 | 5.0 |
| Channels required | 126 | 144 | 162 | 180 | 180 | 180 |
| Terminals required | 252 | 288 | 324 | 360 | 360 | 360 |
These figures are based upon an effective availability of the computers for interactive use of 60 hours/week; an average processor utilisation of 60%; an average channel utilisation of 60% and two terminals per channel.
This section summarizes the findings of the Software Sub-Group based on the views that were expressed during the visits listed in Section 4.
The major requirement found was for easy-to-use interactive systems with good FORTRAN facilities. There was a strong general recommendation that emphasis should be placed on the ease of use of all software.
In the applications area there was a need for a centre to undertake a co-ordination and documentation role. There was also a very widespread requirement for good standard software in the area of graphics.
The command language of the interactive system should be very easy to use. There is a need for the FORTRAN system to include a library designed to aid interactive use. This should include functions such as command decoding, string handling, menu handling and similar functions. The library should be well documented and well maintained.
There is some demand for an extended version of FORTRAN, as ASA FORTRAN is primitive in a number of respects.
A limited need was found for other languages:
In general the demand for application packages was lukewarm. This may indicate some lack of awareness of what is available but also indicates that it is necessary to make a careful assessment of user demand before putting effort into the support of application packages.
The major requirement in the applications area was for standard graphics software. The demand extended from very basic routines through software of the GINO type to specific application level packages (eg for integrated circuit mask design, printed circuit board layout and similar activities). At the present moment a very wide range of software is in use ranging from home written routines to complex packages supporting specific devices. There is clearly a primary need for some groups to undertake a coordinating and documenting role over the whole range of graphics software.
Other requirements for packages and special purpose languages were:
The strongest area of demand was for the finite element packages.
In the area of on-line experimental work a general purpose interactive statistical package would be of value as would cross-assemblers for the common minis in use (PDP11, Nova, etc).
Users expressed a desire for interactive facilities when engaged in large scale analysis. They wish to prepare and check input data interactively and to be able to inspect output interactively - particularly graphically. Their requirements for central processor power are very large and hence they realise that the main computation cannot be done interactively. Their requirement is that the interactive facilities be connected to the largest available SRC computers (ie the 360/195 at the moment) and that jobs be submitted to the large machine using data prepared in the interactive machine. There is a need for some scheduled period in which priority response can be obtained on the 360/195.
It is very difficult to estimate the central processing power required from SRC's major computers for this type of work. The Technical Group provides the estimate given in Table 5.8 based upon the current use of about a 360/195 for all engineering batch processing. It should be noted that no assessment has been made of normal batch processing power required by engineers.
| 1975 | 1976 | 1977 | 1978 | 1979 | 1980 | |
|---|---|---|---|---|---|---|
| Power required from SRC in 360/195 units | 0.03 | 0.1 | 0.2 | 0.3 | 0.4 | 0.5 |
