At its meeting in Bristol on 14 November 1979, the Interactive Computing Facilities Committee made some major decisions about the future ICF programme at Edinburgh and UMIST. This followed some 12 months of discussions with the two User Committees and with the managements of the two sites. The ICFC has recommended to the Engineering Board that the DEC-10 at UMIST should be replaced by a PRIME 750 and that a package offer from DEC should be accepted. to trade-in both existing DEC-10s (model KIs) for a 512 Kword DEC-10 model KL 1091S which would be installed in Edinburgh. The total cost will be £448,000, but this will result in very considerable savings on recurrent expenditure.
At UMIST, the power of the PRIME 750 installation is expected to be not less than that of the KI-10; the support team have established that the great majority of their present users will be able to transfer to the PRIME. There will be an overlap period of about 12 months with new and old machines both operating. The DECnet node will remain at UMIST for as long as it is required for the connection to Edinburgh. ICFC recognises important advantages in transferring most of UMIST's present users to the same PRIMOS operating system as will be running on 7 other ICF machines by early next year. They will be helped in this transition by the UMIST support team, who will not only become experts on the PRIME 750 but will eventually provide central systems software and user support for all PRIMEs in the ICF network.
The Committee recognises the need for ICF to retain a major DEC-10 centre, to provide: first, software compatibility for a limited number of users with DEC-10-specific programs and second, a programming environment particularly rich in features useful to the Artificial Intelligence community. The KL 1091S is intended to serve some 30-40 concurrent users (instead of the present 25), allowing each an increase in useable program size and available processor power. This machine should cope comfortably with the increasing expectations of these users over a period of 5 years. The user support team at ERCC will be somewhat strengthened; later, some ERCC staff will study the applicability of linked, powerful single-user systems to the needs of Artificial Intelligence and of other ICF users. These systems will probably supplant the central 36-bit machine in the longer term. A key feature of the plan is the retention of the existing DECnet network to allow remote access to those users anywhere in the UK who have an over-riding need for the DEC-10 service. This currently includes nodes at Dundee, Sussex, UMIST and Chilton: a gateway machine at Edinburgh will also provide connection to the SRC's X25 network and so to the GEC, central-batch, and later PRIME facilities.
These proposals still require the approval of the Engineering Board and of the Science Research Council itself, so that no firm timescale can yet be given. Although they provide savings in the longer term, the initial outlay is large and will come at a time of increasing stringency in Government spending. Nevertheless, the ICFC has determined its chosen replacement policy, has given the necessary spending a high priority in relation to its other commitments, and intends if possible to install both new machines during 1980.
This working party presented a report recently to the ICFC. The working party was set up to consider the software needs for both pre-processing and post-processing of material which would be used (or generated) by large central packages. Emphasis was placed on requirements which covered a number of disciplines. The main conclusions of the report were:
The report recommends some 12 man-years of effort over the next 3 years from within the ICF at Rutherford. and some 5 man-years of external assistance. ICFC have accepted the report, and its implementation is now being considered. We hope to recruit two more staff locally, and to identify University groups interested in participating in an extra-mural agreement to tackle one or more of the problems. If anyone is interested in this work. please contact Eric Thomas.
ICF hardware has been through a continuous process of change since the Facility came into existence. This process is now almost complete and in case anyone has become confused by it, a list of ICF mainframes and multi-user minis appears below. New machines which have been approved but not yet installed are indicated by *. Where the ICF upgraded an existing machine this is indicated by #.
The VAX at Leeds has recently replaced a PDP 11/70 previously upgraded by ICF. We hope that all the machines listed above (apart from the UMIST PRIME 750) will be installed before April. This would complete the ICF multi-user mini installation programme, apart from one GEC 4000 series machine which may be installed in the next financial year if funds are available; a short-list of possible sites has been prepared.
The Finite Element Package evaluation exercise described in 'Rapid Response' No.9 is now over. The Finite Element Steering Group considered the results of the evaluation and has selected ASAS and BERSAFE for support under the ICF programme. ASAS is a large, well-documented package with a good user interface. It can handle linear elastic static and dynamic (modal analysis) problems. BERSAFE is smaller and consists of a suite of programs including the following:
Other features are Fourier elements which allow non-axisymmetric loads on axisymmetric structures, and an extensive fracture mechanics facility.
The first release of the Finite Element Library is now available complete with documentation. The FE Library is intended as a tool for FE developers, and consists of two parts, Level 0 and Level 1. Level 0 is very much on NAG Library lines and consists of a set of subroutines which perform most of the standard tasks required in a FE analysis: shape functions, matrix assemblers, equation solvers, eigenvalue/vector extractors and so on. All work is done in main memory at present. The Level 1 programs are essentially examples of how to use the Level 0 Library. Currently, at Release 1, there are four segments:
A large amount of code already exists for other applications but it needs conversion to our format, and documentation needs to be prepared. These applications are:
An FE Library training course was held at Rutherford on 29 October to 2 November. We are considering holding regional courses - please get in touch if you are interested, or if you have any comments about library contents.
A Finite Elements User Meeting was held at Rutherford on Tuesday 18 September. 73 delegates from many Universities and Polytechnics in the UK attended. Talks were given on the ICF; the work of the FE Special Interest Groups; FE Packages supported by ICF; pre- and post-processing of FE analyses; and the FE Library. The meeting ended with an extensive discussion period where users could express their comments on future developments. Interest was expressed in the following:
It was felt that future meetings should be on more specific areas, fluid mechanics, say - rather than Finite Elements in general. A report of the meeting is available on request. Further information can be obtained from Ken Robinson at Rutherford.
Since publication of the last 'Rapid Response', we have seen the departure of 3 people who contributed greatly to setting up the ICF in completely different ways.
All those associated with the Edinburgh DEC-10 know how much Bill Hay contributed as Systems Manager of that machine, before as well as after the time that it was taken over by ICF. He made important contributions to the TOPS10 operating system and was one of the main architects of the DECnet work linking the DEC-10s at Edinburgh, UMIST and Dundee, and the communication node at Chilton. Bill's presence as a technical expert in everything to do with DEC-10 systems, and his continuing sanity at times when everything around him seemed crazy, are things that we are going to miss. We congratulate him on his appointment as Communications Manager at ERCC, and no doubt we will still benefit from his expertise in his new role.
Neil Kennedy was Systems Manager of the Nottingham PRIME 400, another man with an honourable service record from long before the machine was taken over by ICF. His work in support of the SAMMIE project for interactive evaluation of a man and his workplace is well known. He has also contributed greatly to the harmonising of the operating system run on the Nottingham PRIME 400 with that now run on the Chilton PRIME, and to the build-up of use on the Nottingham machine by various new research workers supported by SRC grants. Neil has moved to Compeda Ltd, and as with Bill it is quite possible he may be in a position to do ICF a good turn in the future.
Mike Sterling is known to all of us as the enthusiastic driving force behind the Interdata 8/32 in the Control Engineering Department at Sheffield. For many years he has fought for the cause of better interactive computing for engineers at Sheffield and, as a result, the Control Engineering Department is probably the best provided for in the University. In October 1977 the ICF agreed to take a 45% share in the Interdata in return for the usual enhancements to increase the number of users that it could support simultaneously, and Mike gave us valuable support in assessing that machine in comparison with others being considered as a possible standard multi-user mini. The fact that our final decision was in favour of GECs and PRIMEs means that ICF will gradually be reducing its involvement with the Interdata 8/32, but will be assured that SRC research workers at Sheffield have a very adequate local facility before that happens. It is a pleasure to congratulate Mike on his appointment to a Chair in Control Engineering at Durham. We wish him well and wonder how long it will be before Durham pops up as a node in the SRC network.
The Chairman of this committee is Dr J C Baldwin of University College, Cardiff and the secretary is Mrs J J C Hutchinson of Rutherford. The remaining members are:
The membership listed above includes as ex-officio members the installation managers of those 4000 series machines which have established a significant user population. Each remaining manager will join as and when the user population on the machine for which he is responsible attains a similar level.
Updates to available software:
The PE2D code for solving non-linear 2-dimensional problems in electrostatics, magnetostatics and steady state eddy currents has had two new releases since March 1979. The first of these provides a new system for storing data files and also an option for axisymmetric geometry. The second allows automatic mesh data checking; enhanced graphics; improvement to speeds, and also electron optics facilities in the post-processor. Further development is in progress. Copies of the rewritten User Guide are available.
The BIM2D programme for constant permeability problems now has an enhanced graphics post-processor. A user guide is available in addition to the extensive on-line HELP facility.
New codes:
TLM, the Nottingham University Transmission Line Method code for solving Maxwell's equations, has been implemented on the IBM system. A user guide exists.
A user course was held on December 5 and 6 at Rutherford. It covered the following software packages: GFUN; THESEUS; BIM2D and PE2D. A similar course will take during the spring of 1980 if there is a demand. More information is available from Jim Diserens (Rutherford Laboratory).
Approval has been given for the Unit to install rotational hardware for the Vector General display to complete the upgrade begun in 1978. Delivery is expected shortly and should enhance the facilities of this highly reliable display. The Unit reports that the system is now running under RSX-11M and the improved memory capacity, visual display buffer and large disk system installed a year ago have significantly enhanced the service. Five active groups are already using the service including Professor F Walkden's group at Salford and Professor Brotton's group at UMIST. Further details of the work of the Unit may be obtained from Dr R Hubbold, the Director.
A PDP 11/45 in the Department of Biophysics at Leeds University has been linked to the ICF VAX 11/780. The link operates at 19.2 Kbaud over Tariff-T lines. The high speed is possible because the lines do not pass through any exchange on the way. The DECnet software permits three modes of communication: program-to-program, remote file access and remote device access. The PDP 11/45 supports a high-performance graphics device.
Many research and development workers in this field are now on the DACD group's mailing list and will be sent copies of the DACD Bulletins. These provide information about the comprehensive set of modern analytic software for Electric Circuit Design Systems currently on the ICF. Anyone still requiring a copy should contact Bridget Astridge at Rutherford.
One of SRC's two astronomical observatories, the Royal Observatory Edinburgh, is using the ICF's standard GEC system software on a 4000 series computer. The software, including updates, is maintained by ICF but the service on the machine will be managed by the Observatory for the needs of its research programme.
Our good wishes go to this project, which is a 'mini-ICF' aimed at the astronomy community. It will be managed by staff of the Computing Division at Rutherford, some of whom have been working in ICF up to now. Starlink will have its own newsletter - there is already an astronomical journal called "The Observer", so perhaps we should call it "The Morning Star"?
In the last issue the address at the bottom changed slightly. This time there are two more small changes: one commemorates a previous merger of SRC establishments, the other is caused by a further merger which is going ahead at present.
This issue will reach readers after Christmas, but there is time to wish everyone lots of Allocation Units in the New Year.
