Additional paper relating to item 5 of the agenda for the meeting on 29 January 1987 [CCC(87)4] (Ref Nos CHE/24, CHE/24/3)

Introduction

Last year's report closed with the appointment of the working party on the replacement of the ICL 2988. This year's events have been dominated by the activities of the working party and the preparation of the case submitted to the Computer Board.

The working party began its work in July 1985 and has continued to meet on a regular basis ever since. Its terms of reference were to make recommendations to the Computing Council on the form of the proposals to be made to the Computer Board, the Council having the formal responsibility for drawing up the submission on behalf of the General Board.

Two crucial decisions taken by the working party were to recommend the replacement of the ICL 2988 system by open tender, and to recommend that the available funding should all be devoted to a system from one manufacturer. The first of these decisions required little debate. The Computer Board's declared policy is for major procurements to be conducted by open tender, and it was agreed that even if the final outcome were to be the replacement of the ICL 2988 by a new ICL configuration, the choice should only be made after comparative evaluation of proposals from different manufacturers.

The second decision required much greater deliberation. The present service is divided, roughly half-and-half, between the ICL and the VAX systems and it could be argued that there are advantages in continuing with a dual approach - each type of system carrying the applications for which it is best suited. The counter-argument, which eventually carried the day, was that there can be considerable economy of effort (both for users and for the Computing Service) if there is only one main operating system and user interface to learn and support; this seemed particularly important at a time when ever greater resources are needed to support new activities in distributed computing in departments and colleges. The working party also took the pragmatic view that a large budget might be more attractive to suppliers if offered as a single sum than if spread over more than one type of system.

While these matters were being discussed, a considerable number of visits were paid to user departments, and the evolving model for the submission was also discussed with the Users' Group. Contacts were developed with several manufacturers, some of whom visited Oxford to give presentations on their ranges of products.

The outcome of two terms' concentrated effort by the Working Party was the acceptance by the Computing Council and the General Board of two documents, a medium-term strategic plan for the University and a draft Operational Requirement for the equipment to replace the ICL 2988 system.

These were submitted to the Computer Board at the end of March. The Board considered the University's case at its May meeting, and a Board party then visited Oxford on 21 May. Their conclusion, subsequently confirmed by the June Board meeting, was that the University's strategy should be approved. The Board authorised a very substantial capital sum for the mainframe replacement exercise, an award which was the more warmly appreciated by the University since it was announced almost simultaneously with the retrenchment in UGC funding.

The Operational Requirement was revised in minor respects in the light of comments from the Central Computer and Telecommunications Agency and the Joint Network Team. It was issued to potential suppliers on 20 June, thus formally initiating the procurement process.

This review has so far concentrated on the replacement computer system itself, but the case to the Computer Board also laid emphasis on plans to develop, with University funding, the infrastructure to provide greatly improved access to the proposed new system. The build-up of the communications network is described elsewhere in this report.

A further need identified during the period of the Working Party's activities was for provision of terminals or, preferably, microcomputers capable of terminal emulation in significant numbers in departmental and college locations. This led to the adoption by the Computing Council of a University workstation policy, under which the Council plans to make recommendations of hardware and software which will receive preferential support, and where possible funding through the Research and Equipment Committee. Initially it has been decided that the workstation policy will be based on IBM-compatible hardware, and the choice of a suitable word processing package to be recommended for primary support is to be considered during the coming long vacation.

While the preparation for choosing the next generation of equipment has been a major preoccupation, the maintenance and development of existing services has been continued in the normal fashion. As usual, these are described under succeeding headings.

[Top]

Software and Service Changes

General

One of the main activities of the Computing Service during this year has been the introduction of the Gandalf PACX2000 terminal switching system. The first node, located at the Computing Service, was installed in September 1985 and linked to the existing PACX IV switch. Computer systems and terminal ports were then gradually transferred to it, in a way designed to minimise disruption to users. A remote shelf was installed in the Museum Lodge telephone exchange building in April, and used to support Science Area terminal lines. At the end of the period of this report a second node had been installed and the fibre optic links between the nodes had been validated.

The full configuration envisaged for the end of 1986 is to consist of three nodes (at the Computing Service, the Museum Lodge and the Bodleian Library) linked in a triangle. This configuration thus contains a degree of redundancy, allowing a route from any node to any other node to be maintained even in the unlikely event of a failure of an inter-node link.

To meet the steady growth in demand for national network access, the X25 line connecting the University's packet switch exchange to the Rutherford exchange was upgraded to 48 kbit/s in September 1985.

Work on the FAMULUS77 project was completed by the autumn of 1985, and the software has been handed over to PLU at Edinburgh for distribution and maintenance.

The demand for advice on microcomputer hardware and software continues to grow. For the first time an "open house" microcomputer advisory service was introduced during the year, operating on three afternoons per week. This has proved very popular, and it is hoped to expand this area of activity when staffing resources permit.

ICL 2988

The problems with release 811 of VME which were outstanding at the end of the last report continued into the autumn of 1985. Although ICL believed the problems to be cured, they recurred in September and the site returned to a "Red Alert" state. The last outstanding problem was cleared in November, and since then the VME service has run in a stable and satisfactory manner.

Further developments were made in the use of CAFS. A version of Fortran77 that supported the CAFS Search Option (CSO) was released in October, and CAFS searching was made available from Algol68 in January 1986. New versions of various packages with CAFS support, including SPSS-X, were introduced. A modification to the disc configuration now allows all the FDS640 disc units to be accessed via the CAFS controller.

The terminals still using the MAC service through the 7906 front end processor were transferred to the X25-based NMAC service. The only remaining function of the 7906 was support of XBM printers at six locations around the University. These printers were transferred to either of VAX1 or VAX2, using direct connections and long-line drivers, and the 7906 was decommissioned in June. Use of software already in place to spool output from the 2988 to Cambridge Ring printers enabled this change to take place without affecting the user interface on the 2988 or the VAXs. The only remote printer still driven from the 2988 is a C03 service to the John Radcliffe site.

The build-up of the workload has justified the expense of running the 2988 at weekends, with some operator cover, and this has been done since early in 1986. Various changes have been made to the scheduling software to service the weekend workmix appropriately, and to the accounting software to allow off-peak working to be charged at reduced rates.

A great deal of work has been put into installing and testing coloured book software for file and job transfer. Although the existing ad hoc job submission commands remain available for the present, remote site users will now be encouraged to experiment with submitting their jobs using JTMP or to work by using remote login and file transfer via FTP.

New versions of GIMMS, CLUSTAN, ASPEX, MDSX, SIR, PSTAT, TSP, GLIM, SPSS and Mark 11 of the NAG Fortran Library, were installed. SPSS-X, with CAFS support, superseded SPSS. Version 7 of Ghost-80 has been implemented.

VAX

The main service machines were converted to version 4 of VMS in September. A great deal of effort was put into tuning and monitoring the system in anticipation of performance problems of which other sites had warned, but in practice there was little degradation observable under our somewhat atypical workload.

The disc space on VAX3 was reorganised, with many development files being transferred to a demountable volume, to make more user space available on the permanently mounted disc. This gave more scope for running courses, and for the occasional transfer from VAX1 and VAX2 of users with a requirement for large amounts of low-priority batch time.

A change was made to the login procedures to recognise the type of terminal being used. This removed many problems caused by terminals locking up when sent incorrect control sequences, a problem which had become more pronounced under VMS version 4.

FTP reached a fairly stable state during this year, but the introduction of JTMP into full-scale service led to problems affecting both FTP and JTMP. DEC could offer no immediate, or even long-term, cure for this, and could only recommend reducing the level of FTP concurrency far below that demanded by our workload. Attempts to provide a JTMP service for National Centre users who would prefer to submit jobs directly from the VAX, rather than through the 2988 as in the past, have thus been frustrated for the time being.

New versions of SAS, TSP, PSTAT, REDUCE, GLIM and MINITAB, and Mark 11 of the NAG Fortran Library, were installed. Three new packages, SIMAN, MAPLE and FSIS were introduced, and version 7 of Ghost-80 was implemented. MACSYMA ceased to work with the introduction of VMS version 4, and had to be withdrawn.

[Top]

User Liaison

Allocations

All users were required to complete new registration forms, the wording on which had been revised to take account of the Data Protection Act. This exercise provided an opportunity to remove a certain amount of dead material from the user registration system.

Agreement has been reached to extend the group allocations system to Oriental Studies, Geography, Biochemistry, Psychiatry and the Childhood Cancer Research Group. The former Forestry and Agricultural Sciences groups have been merged, as have the Politics and Sociology groups.

Courses and Seminars

The range of regular courses on the mainframe systems has been expanded this year. In particular, new VAX users are now introduced to EDT rather than to ECCE as the principal editor. Brief details of these courses are as follows:

In addition, seminars and lectures on a variety of topics were presented. A series of 9 lectures in Michaelmas Term dealt with statistical computing, including an introduction to the packages SPSS-X, Minitab and BMDP. In Hilary Term four seminars on database topics were given, two on mainframe facilities (including CAFS) and two on database software for microcomputers. Finally, in Trinity Term there was a series of seminars on assorted topics including graphics, use of JANET, use of the Lasercomp, and SASS-X for beginners.

Documentation

A list of current user guides is given at Appendix B.

Remote Sites

In general, services remained stable during the year. The main topic of interest was the discussion, following the publication of the Forty Report on advanced research computing, of the new facilities which are to become available at the National Centres. As well as the increase in vector processing capacity expected at ULCC and the Rutherford Appleton Laboratory, the arrival of new IBM-compatible front end systems at ULCC and UMRCC will be welcomed; the present Amdahl 470V/8 at ULCC is especially heavily loaded.

The CDC 7600 service at UMRCC is still heavily utilised by a number of Oxford users, and since these machines must be close to the end of their useful lives the continued existence of this service - and the nature of its replacement, if any - is currently of some concern.

[Top]

Computing in the Arts

Lasercomp

During this period the software has been relatively stable, and the workload has continued to grow. The laserprinter proofing facility was brought into service in April 1986 after a number of problems had been ironed out. Initially no charge is being made for laserprinter output, and this has led to a dramatic increase in usage.

Two meetings of external user representatives have been held, and two issues of Lasercomp News published. Following a ruling by the CVCP, external users' project codes were changed from 1 April to indicate that VAT is now charged on all such work.

MICROSET, a program which paginates Lasercheck files, has been made available at Oxford on a BBC micro system in the User Area.

Software development for non-standard alphabets has continued. A program for typesetting Egyptian hieroglyphs is now operational, using Gardiner's hieroglyphic font which was originally digitised for the Lasercomp by Oxford University Press.

New documentation has included a Lasercheck tutorial guide and a new version of the user guide to using the Lasercomp from remote sites. A seminar on publishing scholarly texts was given in Trinity Term 1986. In addition, staff participated in the BCS Electronic Publishing Group, and took part in a workshop on typesetting in Paris in November 1985. Lectures on the Lasercomp and KDEM were given at the University of Manchester, and to the conference of University Registrars at Cambridge.

Kurzweil Data Entry Machine

The KDEM service now has a staff of four, of whom two occupy established posts and two are funded by income from the service.

During the year 44 jobs were scanned. Some of these consisted of several volumes: the longest job was 115 hours, and the average length was just over 15 hours. Whereas originally output was normally sent to the user on magnetic tape, almost half the scanned texts are now sent on floppy discs and a few directly via JANET.

A paper on the KDEM was published in Literary and Linguistic Computing.

Oxford Concordance Program

36 program issues were made of which 11 were in the UK, 9 in the United States, 4 in Canada, 2 each in West Germany, Italy, Australia and Japan, and one each in Austria, South Africa, Norway and Thailand.

In September 1985 work began on version 2. For this new version the program is being completely rewritten in Fortran77. Version 2 will provide the same facilities as version 1, with some rationalisation of the commands. By the end of the period of report, the code was almost ready for user tests. A paper on version 2 was presented at the ALLC conference in Norwich. Two seminars on OCP were also given in Pisa.

Oxford Text Archive

A paper describing the Archive's facilities and outlining plans to expand its activities using CAFS was presented at a symposium in Nice in June, and subsequently published (From Archive to Database, in Brunet, L. ed, Methodes quantitatives en linguistique, Slatkine, 1986). A similar presentation was given at an international workshop on interdisciplinary sourcebanks in the historical disciplines, held in Göttingen in July. An invited lecture was given at the Instituto di Linguistica Computazionale in Pisa in September.

About fifty new texts were deposited. Most of these were produced by the KDEM service. Three important exceptions were: Dante's La Divine Commedia, which was deposited by the Instituto di Linguistica Computazionale in exchange for the works of Shakespeare; the CATSS database (parallel word-aligned texts of the Greek and Hebrew versions of the Septuagint) deposited by Prof. E. Tov; Roger Mitton's enhanced "computer-usable" version of the Oxford Advanced Learners Dictionary of Current English. The last text in particular has proved very useful to researchers from a wide variety of disciplines.

Over the year nearly two hundred texts were issued; of these about one third each went to scholars at American and British universities, the remainder going to countries around the world, including Japan, Australia and Finland.

Text searching software which uses the power of CAFS to search through large bodies of text was further developed. One user to benefit particularly from this was the OUP Shakespeare Department, who used it extensively during the preparation of their new one volume edition of the complete works. The Shakespeare Database was presented to Royalty by ICL (when CAFS received a Queens Award for Industry), but was not, despite press reports to the contrary, in fact much help in establishing the authorship of "Shall I die".

General

The Computing Service has participated in the experimental humanities computing bulletin board set up by the Office for Humanities Communication at Leicester.

An invited paper on the Oxford facilities for computing in the humanities was presented at a conference in Toronto in April 1986. Seminars and lectures were also given at Thames Polytechnic, Durham University, the Norwegian Computing Centre for the Humanities, Bergen University, and the Universities of Waterloo, Saskatchewan and Calgary.

[Top]

Performance and Reliability

Mainframe services

Interruptions to service have in the main been attributable to software problems rather than hardware faults. The problems with VME release 811 on the ICL 2988 and with communications software on the VAXs have already been described above. Apart from these, there have been a few incidents involving fixed disc units on both systems, with an FDS640 being replaced in July 1985 on the 2988. Memory errors on VAX1 and VAX2 were troublesome for a period in August 1985, but these were eradicated and have not recurred.

Other equipment

The fire alarm system gave trouble at two points during the year, but the most recent repair appears to have cured the problem and it is hoped that there will be no further incidents of this kind. Two days' service were lost owing to the escape of water into the machine room underfloor void during the Christmas-New Year closedown period. A monitoring system is to be installed in the machine rooms to detect environmental failures of this nature at an earlier stage.

The intermittent fault on the Lasercomp tape deck was still present at the start of the year, but this has finally been overcome. The Lasercomp service also suffered minor but unfortunate interruptions, firstly because of accidental damage to the film processor and secondly through the failure simultaneously of two independent suppliers to deliver new stocks of bromide paper on time.

[Top]

Appendix A. Staff in post - June 1986

[Top]

Appendix B. Current User Guides

[Top]

Table 1. Machine Usage

[Top]

Table 2. Reliability and Mean Time Between Failures

[Top]

Table 3. ULCC Usage

[Top]

Table 4. UMRCC Usage

[Top]