There have been some changes in membership during 1978. Professor A G J MacFarlane of Cambridge University is now Chairman and the other members are as follows:
Dr C J Pavelin of the Rutherford Laboratory is secretary and Dr K D J Root, of the Computer Board for Research Councils and the Universities, attends as assessor.
A GEC 4070 has been installed in the Department of Civil Engineering, Bristol University, and a service is being run. Readers interested in further information should contact Dr J W Smith there. The 4070 for Cambridge University Engineering Department has also been installed and is also running a service; it will shortly be equipped with an IMLAC PDS-4 high-performance refresh graphics display. We expect that the last of the 4070s approved in the the summer of 1978 for Glasgow University, Faculty of Engineering will be installed in March.
Two more 4070s are to be installed during April or May at Newcastle University (Faculty of Engineering) and Cranfield Institute of Technology (Department of Computer Studies).
Sites for a further five minis were selected in January for installation before March 1980.
They are:
The PRIME 300 at the University of East Anglia (School of Computing Studies) is being upgraded to a PRIME 400. A major project to be supported is Dr A R Forrest's work on software for an algorithmic approach to the processing of geometrical data.
The Special Interest Group for Digital Circuit Design (SIGDD) has completed its initial assignment. Its proposals have been approved by the Interactive Computing Facilities Committee, apart from those concerning microprocessors; reactions to the latter are awaited from other Government-supported organisations concerned with providing microprocessor design facilities before final decisions are made. The proposals which have been accepted are set out below.
Facilities for this purpose should be based on the GAELIC program package (see 'Rapid Response' no. 6, section 4). Considerable enhancements are needed for printed-circuit board work - or, as a possible alternative, the software from the Lawrence Livermore Laboratory in the USA should be examined. Detailed proposals should be drawn up for the design of a special camera for the FR80 microfilm recorder (see Rapid Response' no. 5, section 5). This would provide adequate resolution for many masking operations and for the verification of LSI (Large-Scale Integration) sub-structures.
Three large general-purpose circuit analysis packages ASTAP, SPICE and NAP-2 - have been mounted on the 360/195s at Chilton for use by other circuit design groups. They will also be used to check the behaviour of digital systems at the design stage before mask-making. Some manpower should be provided for maintenance, user support and further interfacing.
Software for mapping geometrical specifications of digital circuits on to equivalent analogue circuits is still required.
A device and materials library should be maintained in the PRIME 400, as should information about the available facilities. The PRIME data-base management system should be assessed for suitability as a project control medium for long-term LSI and VLSI work. This facility should be shared with users of applications software in many other disciplines.
The proposals for microprocessors and associated LSI devices were as follows:
It is unlikely that software for test generation and fault simulation can be mounted in the near future. The possible packages are either very expensive or incompatible with our computers. Enquiries will continue.
The SIG has had permission from the ICFC for a report amplifying the proposals to be circulated widely outside SRC. Copies may be obtained from Mrs Cynthia Davis at the Rutherford Laboratory. Comments on the report would be welcome.
One of the additional staff needed for work on digital systems software has been recruited. A start has been made on mounting a library of software for Transistor-Transistor Logic (TTL) systems.
One essential element in the plans is a library of microprocessor software including cross-assemblers, simulators etc. Such a library has been set up on the UMIST DEC-10 by staff of the Microprocessor Unit under Dr E Powner in the Department of Electrical Engineering and Electronics at UMIST. Documentation and user information are being compiled. Documented software donated by users for inclusion in the library would be welcomed.
A Special Interest Group in Control Engineering (SIGCE) has been set up. The Chairman is Professor D Q Mayne of Imperial College and the Secretary is Dr M J Denham of Kingston Polytechnic. The other members are:
An open meeting was held in September. Since then the SIG has decided that a library of specialised subroutines should be set up. A survey of control engineering software is being carried out; with the survey forms a questionnaire was circulated asking the respondents what software is needed in future.
The next open meeting will be held in Cambridge on 25 April; details will be distributed nearer to the date. Our present mailing list may not cover all potential users of the software. Any interested reader working in control engineering should contact one of the SIG members to check that he is on the list.
An open meeting was held in Edinburgh on 25 October 1978, chaired by the SIG Chairman, Professor T W Maver of Strathclyde University. Short talks were given on the present state of the ICF network, plans for its extension, and work on response time measurement on the Edinburgh DEC-10s. Miss J Frith, a member of the SIG, described the results of a survey of low-cost hard-copy graphics terminals.
The results of a recent survey of CAAD users were described - the replies showed that many groups are working in isolation and a method is needed to inform research workers about the available software. The meeting agreed that a register of software, and more information about SRC-sponsored research in this field, are both needed.
After a discussion on data-bases the meeting recommended the investigation of various data-base management systems for possible mounting on ICF machines. One particular candidate is the BOS system, part of the OXSYS computer-aided building system.
(NOTE: Since the meeting the Interactive Computing Facility has decided to investigate BOS but the work is held up at the moment for lack of manpower for detailed evaluation )
Various demonstrations followed, mainly of graphics systems. One was of the results of a feasibility study commissioned by the SIG, which used a microprocessor to provide local intelligence for a standard Tektronix terminal so as to reduce the demands on line speed and reliability made by graphics work over remote links.
The Chairman of the SIG proposed a five-year programme of work which was accepted by the November meeting of the ICFC. A total of 15 man-years of effort will be specifically devoted to SIG work, and another 5 man-years of general Interactive Facility effort (eg for user support) will also be needed. The main aim is to develop a replacement for the GFUN package, to provide a new solver for Poisson's equation. The solver will be adaptable for a wide range of applications; separate specialised pre- and post-processors will be provided to make it convenient to use in magnetostatics and electrostatics. A package for eddy current work is also being developed. More details will be published later.>
A five-year programme of work has been approved by the ICFC. Some facilities will be shared with users of Digital Systems software as described above. A detailed report will appear in the next issue.
The PRIME 400 at Chilton was initially one of the two multi-user minis ("MUMs") chosen for development and assessment for suitability as the Interactive Facility standard. When the collaboration with three universities - which formed the major part of the assessment - was complete, the PRIME was upgraded to support a user population of at least 100 researchers (10 simultaneous users). The configuration was also designed to support investigations into mounting large-scale scientific data-bases on such machines.
The current configuration is as follows:
There is a proposal from the Electrical and Systems Engineering Committee of SRC's Engineering Board to make a substantial upgrade to the PRIME including an additional processor. This would be mainly for users of the SRC's Electron Beam Lithography Facility, but the upgraded machine would still be wholly under ICF management.
The PRIME is linked to the 360/195s, permitting remote job entry from the PRIME filestore with results returning to the filestore if required. There are also asynchronous links to the PRIME 400 at Nottingham University and to the Chilton GEC 4070, running File Transfer Protocol. External users currently access the PRIME asynchronously via dial-up (1200/75 or 300 bps) or leased Post Office lines, X25 software and hardware should be available shortly.
The PRIME runs under the PRIMOS 4 operating system. FORTRAN is the main language and BASIC is also provided. Many other languages are available, or becoming so, on PRIME computers and any requests are considered seriously. A preprocessor for structured FORTRAN is being implemented as part of a package of software tools.
A CODASYL database management system (DBMS) is being evaluated and the context retrieval package STATUS has been purchased. GINO and FINGS are the standard graphics packages and the NAG library has been mounted.
The user population is already about 100 and although the configuration was intended to support 10 simultaneous logged-in users, it is common to have up to 15. To keep down the load, internal ports have been restricted so that Rutherford staff working on systems or applications find access more difficult than external users (a fact resented by the former!). Allocation units are being issued but until recently have not been controlled; the recent introduction of controls should improve matters.
A programme of performance measurement began in mid-year and will continue. The response time is normally quite adequate with 10 users, but with 14 terminals in use responses can take several seconds. The machine is quite heavily loaded during day-time; the applications programs using most resources are those in electromagnetics.
Up to now, user support has been limited to the generation of documentation and correspondence through the PRIME filestore, but now one staff member, Mr M T Keane, has been appointed for full-time user support. A user committee will be formed shortly.
The basis for the accounting systems used on the DEC-10s was described in 'Rapid Response' nos 3 and 4. There have been some minor changes since then and systems similar in principle have been introduced on the PRIME 400 and GEC 4070 computers, including those 4070s which are being installed in Universities. The DEC-10 systems are now being modified to take account of new information about the capacity of the machines. This article summarises the arrangements now approved for all ICF computers.
The basic unit of computer use is the Computer Resource Unit (CRU). This is a weighted sum of the CPU time (seconds), connect time (minutes), I/O transfers and (on the DEC-10s only) core occupation. The I/O measures are of different quantities in different units on the various computers. The number of CRUs is converted to Allocation Units (AUs) by multiplication by one conversion factor determined by the model of loading for the particular machine concerned, and by another factor which can have values between 0.1 and 1.0 depending on the time of day at which the machine is used. This second factor encourages use of the computers when they are lightly loaded; a high ("peak") value applies to weekday use during day-time, an intermediate ("standard") value applies to use during weekday evenings, and a low ("discount") value to use at night and at weekends. The exact levels of the factor used, and the details of the periods during which they apply, vary between the different machines. The aim is that an AU should represent the amount of computing resource used in one week by an average user; the latter would use on average 0.3 AUs during one hour's connect time at peak periods, if he were running jobs interactively.
All batch jobs attract the "discount" conversion factor, irrespective of the time of day or week at which they are run.
The resource limiting the load on the DEC-10s is CPU time; 62% of the total can be made available to users during peak periods, which is less than originally estimated. More can only be provided by letting response-time rise to an unacceptable level. The weighting factors used in calculating CRUs on the DEC-10s have been changed accordingly.
SRC-supported research workers have ICF resources allocated in AUs; so do those who access the facilities at the discretion of the Director of the Rutherford Laboratory (mostly for "pump-priming"). Certain classes of organisation pay to use the facilities.
Special measures will be taken to ensure reasonable treatment of existing users whose grants are affected by the change in the value of an AU.
All ICF machines have rationing schemes; that on the PRIME 400 has only just been put into effect, while on the GEC 4070s use is being monitored but not controlled as yet. All the rationing schemes except that at Edinburgh are based on the idea of dividing up the project's total AU allocation into equal rations for each 4-week statistical period. Only a limited number of unused AUs from one period can be carried forward into the next. The UMIST scheme is based on the WARDEN system developed at Essex University.
The upgrade of the facilities of this unit is now complete. The enhanced service is running and can support several simultaneous users. The ICF has invested £36,000 in the upgrade, mainly for hardware, and is funding one staff member to work on user support for three years. A dial-up link is available to the UMIST DEC-10; work is in progress on a link via a 1eased line.
A working party has been set up to consider the need for computer-aided design software running interactively on multi-user minis and providing efficient links with major batch-processing packages. We hope that the same pre- and post-processing software (eg geometrical modelling) will frequently be applicable across a range of disciplines. The working party is chaired by the Head of the Interactive Computing Facility, Mr Davey, and consists of one representative from each Special Interest Group with Mr M J Newman of the Rutherford Laboratory (Technology Division) as secretary. Two meetings have been held so far and copies of the minutes are available from Mrs Cynthia Davis at the Rutherford Laboratory (see section 2, above).
This link, from Edinburgh to the Rutherford Laboratory, is now working. It uses software from the US National Institutes of Health, which has been modified by staff of the Edinburgh DEC-10 installation. Some of the software resides in the DEC-10 and some in the PDP-11 front-end, which emulates an IBM 2780 work-station.
Jobs can be spooled on the DEC-10 for submission to the 360/195s and the output can be routed back to the DEC-10 filestore, but it is not possible for DEC-10 users to interrogate the 360/195s. The UMIST DEC-10 communicates with the 360/195s via Edinburgh.
The system managers of the two DEC-10s have asked, through their User Committees, for a few volunteer users to act as "guinea-pigs" to try out the link; such volunteers would be a great help in checking and improving its performance. Longer-term plans to provide full networking between the DEC-10s and other machines, using protocols based on the Post Office's planned X25 system, are being implemented for the ICF by a group at York University.
Investigations have begun into the future of the DEC-10s. Despite the upgrades completed early in 1977, parts of these computers are seven years old and the load on them is growing; the Edinburgh machine in particular is very heavily loaded. The investigations should be complete by September 1979.
In addition to summaries of the software proposals for electric circuit design and electromagnetics, I hope to include a description of recent work on finite elements software.
