This issue gives details about the organisation of the Interactive Computing Facility and the major installations and network plans. It also reports on the users' software meetings which have been held recently.
The first issue outlined the report of the Technical Group on Interactive Computing. Among other things the report recommended that the DEC-10 computers at the University of Manchester Institute of Science and Technology (UMIST) and the University of Edinburgh, which were originally provided for particular research applications needing a great deal of interactive computing, should be upgraded to provide a general interactive computing service. The enhancements are now complete; more detail is given below.
This DEC-10 computer was originally provided by SRC for Edinburgh University for research on computer-aided design; it is housed in the Edinburgh Regional Computing Centre.
The operation of the machine is controlled by the TOPS-10 monitor. The following DEC-supported compilers are available: ALGOL, BASIC and F10 FORTRAN. A macro assembler, MACRO10, and a linking loader with overlay facilities, LINK10, are also provided.
Compilers for APL, BCPL, SIMULA and IMP are available. GINO Mark 1 is being used and will shortly be replaced by the Mark II version. TEKPAK Mark I, a simple package for driving Tektronix terminals, is also mounted. The NAG library of numerical subroutines is being obtained.
The Computer Science Department's Computer-Aided Design Group, in collaboration with the Wolfson Microelectronics Liaison Unit, have produced GAELIC, a general-purpose draughting aid for electronic design.
The Department of Artificial Intelligence has a number of specialist compilers and packages. Major support is given to POP-2, which is being developed to give compatibility with POP-10. Three dialects of LISP are available, including MACLISP from MIT; CONNIVER, a package written in MACLISP, can be used to produce and manipulate complex data bases.
As at Edinburgh, the TOPS-10 monitor is used and the same set of DEC-supported compilers is available.
The range of software reflects the machine's previous main purpose of serving the Control Systems Centre where it is housed. The major system written by the Centre is the Computer Aided Control Systems Design Suite. This provides on-line analysis, design and simulation of single-input/single-output and multi-input/multi-output control systems. The suite consists of a supervisor package controlling various applications packages. All the programs can be used in conversational mode and graphics is used extensively for the presentation of results. A separate device-independent graphics package provides simple graphics facilities for FORTRAN users.
The Rutherford Laboratory is responsible for the management of the UMIST and Edinburgh installations. Contracts have been placed with the two institutions for the operation of the facilities. A series of meetings has been set up, involving the operating staff at Edinburgh and UMIST and Rutherford Laboratory staff, to monitor the implementation of the contract. Matters of general policy will be referred to the Interactive Computing Facilities Committee (ICFC) of SRC's Engineering Board. The ICFC will also receive reports on the operation and utilisation of the facilities.
A user advisory committee already exists at Edinburgh under the chairmanship of Dr J A M Howe and one is being set up at UMIST under the chairmanship of Mr Owen Mills. These committees will advise the Rutherford Laboratory and the ICFC on the operation and development of the two installations. Both chairmen are members of the ICFC.
The previous issue outlined the original proposal to assess the suitability of various minis for multi-user interactive working, in collaboration with a number of universities. Two different machines - the PRIME 400 and the GEC 4070 - have been bought after an extensive tender exercise. The PRIME 400 will stay at the Rutherford Laboratory to provide a service. The GEC 4070 will be moved into a university department, after development work has been completed, for assessment in a user environment.
Initially, working FORTRAN programs with little or no assembler content will be mounted on both machines; later, more complex programs will be mounted to test each machine to its limits. The original authors of the programs will do this work in most cases. The Rutherford Laboratory staff will concentrate on providing library routines (eg graphics libraries), measuring performance, enhancing the operating system and providing systems documentation. The two machines will also be linked to each other and to the 360/195.
The initial set of collaborators and topics is:
Other university departments will probably be invited to assist in the collaboration as the requirements for test programs increase in complexity. If each program can be mounted on both machines, a realistic comparison can be made.
A separate exercise, to upgrade four mini-computers in universities to enhanced multi-user status, is going on in parallel. More details will be given in the next issue. Between them, the two programmes should give experience of multi-user interactive working on a wide range of minis from different manufacturers.
The ICFC has also recommended that one person should assess single-user minis, eg the Tektronix 4051, IBM 5100, Alpha-16 etc, working closely with university groups.
The machine was delivered to the Rutherford Laboratory in December 1976 and was fully operational within a few hours of installation. It has since been used heavily and the hardware has proved very reliable. The 400 is the newest PRIME computer. It differs from the well-established PRIME 300 in its greater power and extra instructions and facilities. It has 2 Kbytes of 80-nanosecond cache store and 192K of main memory; its power is greater than one Atlas. Other features are 16-bit word length, segmentation and paging and virtual memory up to 512 Mbytes. The machine ordered has two 80 Mbyte disk drives, one 9-track 800-bpi tape drive and both synchronous and asynchronous communication capacity.
The operating system (PRIMOS IV) does not take full advantage of all the features of the hardware and there is great potential for improvement in the future. The configuration will give acceptable response to six intensive interactive users simultaneously; the central processor is probably powerful enough to support ten users. We hope to begin a general user service just after midsummer.
The 4070 was installed on 16 February 1977. It is a slightly slower version of the well-established GEC 4080 and its power is less than one Atlas. The configuration includes 192 Kbytes of store and two disk drives. It has a very sophisticated operating system based on many autonomous processes; the hardware does many things such as process switching which are done by software on other systems. There have been difficulties in running FORTRAN programs which use large arrays; a new FORTRAN compiler should alleviate this at the cost of some loss of efficiency.
Several Post Office leased lines are being installed. The UMIST-Chilton and Edinburgh-Chilton links will operate at 9.6 Kbits per second. In the short term this bandwidth is to be divided into four 2.4 Kbits per second sub-channels; the sub-channels will be used for onward-linking other sites which will have leased connections to Chilton - among them are Imperial College, Swansea University, Cranfield Institute of Technology, Kingston Polytechnic and Southampton University.
The Edinburgh site has a leased line to Sussex University and there is to be a link to the ABACUS project at Strathclyde University. UMIST has leased lines to Bradford and Sheffield Universities; one to Leeds University is being installed and one to Liverpool Polytechnic is under consideration. There are many Tariff-T lines to the main sites. These links have been provided for various purposes over a number of years; they are now being rationalised so that they can be used in the most economical way, eg a communications concentrator may be installed in the London area to reduce telephone costs for the large number of users there. Responsibility for lines and modems is being transferred to the main sites rather than residing with the users. Both Edinburgh and UMIST have dial-up capacity: UMIST has four Datel 200 and four Datel 600 lines. Edinburgh's capacity is to be brought up to the same level.
Connecting the 360/195 at Chilton to the DEC-10s has been delayed by technical difficulties; the original choice for a gateway machine has been dropped and alternatives are being considered. In the long term the multi-user minis at Chilton will also be connected to the DEC-10s though similar technical problems will have to be solved first.
The Technical Group Report did not define software needs. The ICFC has recommended that up to 7 people and £80,000 should be available for software in 1977/78; it decided to consult potential users about the best way to use these resources and about what would be required in future years.
A series of one-day meetings was held at the Rutherford Laboratory; potential users of interactive computing were invited. The different meetings were for different subject areas. Each meeting began with a description of the Interactive Computing Facility by a member of the Rutherford Laboratory staff. There followed a talk on what software is already available and then talks on various specialist topics, each followed by a discussion. A final discussion attempted to define software requirements. Here are short reports on the individual meetings:
33 research workers from external institutes and 15 members of the Rutherford Laboratory staff attended. Professor F Walkden of Salford University chaired the meeting; Dr G Manning outlined the Interactive Computing Facility and Dr K Robinson (Rutherford Laboratory) described the software already available.
Dr R K Livesley (Cambridge University) spoke on software requirements for university research workers. In discussion, it became clear that many university staff are unfamiliar with the available packages. The meeting did not recommend specific packages, but defined criteria by which they should be selected. Software should be available on the 360/195 at the Rutherford Laboratory and should be provided on the interactive machines for data preparation and the examination of results. Five topics of special interest were identified and groups may be set up to coordinate research into these topics:
It was recommended that the Rutherford Laboratory should coordinate the work of the groups and provide administrative effort.
Purchased software will probably be needed for non-linear analysis. Several packages for linear static analysis are being assessed. eg ASAS, BERSAFE, PAFEC and FINESSE. Two programmers will be needed to support packages and special interest groups.
39 research workers from 19 institutes attended. The meeting was chaired by Professor P J Lawrenson (Leeds University) and Mr J E Hailstone described the Interactive Computing Facility. Mr C W Trowbridge (Rutherford Laboratory) described the existing software.
In magnetostatics the software is comprehensive. The main program is GFUN3D, developed at the Rutherford Laboratory; it can be made generally available if more manpower is provided for further development. Other suitable software in this area would need to be bought from outside and this will be considered by a special interest group.
In other areas the software developed by users of the facility should be made available to other users.
A special interest group was set up; initially the members will be:
The following likely fields for immediate development were identified in addition to GFUN3D:
A way should be found to make the results of the work available to industry. The work of research students needs co-ordination so that a useful range of software is built up from their efforts.
29 research workers from external institutes and 4 members of the Rutherford Laboratory staff were present. The meeting was chaired by Dr Howe of Edinburgh University. It opened with a description of the Interactive Computing Facility by Dr Manning. Mr R Rae described current software at Edinburgh for Artificial Intelligence research. Short talks were given on specialist areas - vision, natural language, computation theory and theorem proving.
The meeting recommended that the ICFC should provide at least two systems programmers for AI work. Their main function would be to implement and support software in general use. (NOTE: Two systems programmers are maintaining continuing software support at Edinburgh until a policy has been determined). The meeting also recommended that some systems programmer time should be reserved for user support.
Two meetings were held in February. The first dealt with digital circuits and the second with other types of circuit. A report will be included in the next issue.
Dr A Sloman of the Cognitive Studies Programme, School of Social Sciences, University of Sussex, makes several separate suggestions.
Dr A M Andrew of the Department of Engineering and Cybernetics, University of Reading, writes about the use of LISP for Artificial Intelligence Research, reflecting on the discussion at the AI users' software meeting: "Workers in the AI field have some problems because LISP is not usually implemented on computers of non-American origin." "I do not have experience of the use of LISP and may be underestimating its sophistication, but it seems to me it would be fairly easy to write an ALGOL 68 (or ALGOL 68-R) program which would run LISP programs. The LISP program and its data structures could be represented as tree structures in ALGOL 68 and the system would have the essential LISP feature of being able to operate on the program as if it were data. The usual objection to schemes which use one language to interpret another is slowness of operation but I do not think the time penalty would be anything like as serious in this case as in other interpretive schemes. The scheme would allow LISP to be used on any computer on which ALGOL 68 is implemented." "I would be interested to know what LISP users think of this idea. If there is some insuperable difficulty I have not seen (as there may well be) I would be glad to have it pointed out."
The first issue was written by Anton Walter; he devised the production system using hardcopy output from the FR80 microfilm recorder as the original for the production of printing plates. I have now taken over from him. In the next issue I hope to cover:
Our circulation list stands now at over 800 copies. Anyone not yet on the circulation list who wants to receive future copies should write to me at the address below.
