Annual Report

July 2nd 1984 - June 30th 1985

Oxford University Computing Service


Introduction
Software and Service Changes
User Liaison
Computing in the Arts
Performance and Reliability
Appendix A. The ICL 2988 Replacement Working Party
Appendix B. A Report on CAFS Usage at Oxford
Appendix C. Current User Guides

Table 1. Machine Usage
Table 2. Reliability and MTBF
Table 3. ULCC Usage
Table 4. UMRCC Usage


Introduction

This year saw the installation of the remaining equipment purchased from funds made available by the Computer Board as part of the mid-term review. The 4800 bps terminal service to the 2988 started in October 1984, and extra disc storage was added to the VAXs in November. The two main VAXs were upgraded to 11/785 processors at the beginning of March.

This last change involved considerable problems. CCTA had been notified of a deficiency in the floating-point performance of early 11/785 processors in January, but knowledge of this fault within DEC seemed almost non-existent, and the Service only managed to obtain confirmation that it had not yet been corrected two days before the first planned installation date. After applying considerable pressure to DEC the Service managed to obtain corrected hardware, which was installed one month later. Service staff gave advice and assistance with benchmarking to other university sites who had installed or were about to install VAX 11/785s; one of these was found to be running, unknowingly, an 11/785 which ran some user programs more slowly than an 11/780.

The demand for advice and assistance with microcomputers continues to grow, and thought is being given to providing extra staff effort in this area. The microprocessor consultancy also continues to expand; an extra member of staff was appointed in October 1984. This service is hampered by the restricted office and workshop space it has to occupy, and planned building works in this area are still awaited. The Computing Service has been much involved in the discussions over the new telephone network. The decision by the University that the rewiring operation should include provision for a data link to any telephone location provides a major opportunity which the Service is eager to grasp. Access to computer facilities can in future be more readily provided to Senior Members and research workers, whether they work in their department or their college, and the means to improve student access will also be available.

The necessary data infrastructure to support these terminal links has been the subject of much research and planning. The data network will be used to access not only the central machines, but also departmental machines and the machines provided for the library automation project which will show its first fruits this year. There is also ever-increasing use of facilities at other universities and research establishments, made possible by the implementation of nationally agreed protocols and the growth of the JANET network.

After careful consideration of the available systems, and discussions with JNT and other interested parties, the Service decided that in order to provide the required connectivity it was necessary to continue the development of its X25 network, but also to front-end the X25 network with a set of linked asynchronous switches. The main nodes of this system will be sited at the Computing Service and at the locations of the new telephone exchanges, with terminal ports at various sites throughout the University. The equipment chosen is the new Gandalf PACX 2000 system; the first node has been ordered and will be delivered in August.

Mention should be made of the Data Protection Act. The effects that this will have on the Service itself are expected to be relatively minor, being mainly in the establishment of improved procedures to ensure the integrity of users’ data and physical security of the machine-room areas. The direct effects on the users may be greater. The reply from the Data Protection Registrar to the University’s proposals for registration of all its research-related activities is awaited with interest.

In 1987/8 the ICL 2900 equipment will reach the end of its allotted ten-year life-span. The long process of choosing its replacement has begun, with the appointment of a working party to advise the Computing Council on the form of the bid to the Computer Board for the replacement of the 2988. The membership of the working party is given in Appendix A. It is the working party’s intention that as many departments as possible should be visited during the first half of the coming academic year, probably by a small group of Service staff, for discussion of the facilities they would like to see in the next generation of computing equipment.

[Top]

Software and Service Changes

General

The Cambridge Ring has been extended to the Institute of Economics and Statistics in the St. Cross Building, and a printer has been installed there. One of the most interesting additions to the Ring has been an EPS 1200 laserprinter situated in the User Area. This was purchased in December, after a successful trial during September with a loaned printer. The printer has a resolution of 300 dots per inch, and provides high-quality output with a variety of fonts including a Greek/mathematical symbol font. It has built-in formatting capabilities, actioned by control sequences in the text. Output is available in both landscape and portrait mode. Use of this device has built up steadily since its introduction.

Because of further delays with the implementation of standardised job transfer software throughout the academic community, it was decided to transfer the remote services connection to ULCC to the HASP route already used by work destined for UMRCC. As the HASP board is capable of connecting to only one site at a time, a piece of hardware was designed and constructed to transfer the link from one site to another under software control. The implementation of this route enabled the ageing Modular One computer to be taken out of service.

Kermit (not an acronym!), an asynchronous file transfer protocol system, was acquired from Columbia University. Kermit is a well-tried system available on a wide range of machines, and is available free from licensing restrictions. The Service believes it will offer a valuable enhancement to the asynchronous terminal network for the next few years, until standardised protocols become widely available. The Service has acquired or implemented versions on the VAXs and on RML 380Z, IBM PC, Apple and BBC microcomputers.

An expansion of activities in the Department of Atmospheric Physics caused the reclamation of the space hitherto used by the Service to provide a Science Area data centre, housing terminals, graphics equipment and a remote lineprinter. After a number of possibilities had been explored, alternative accommodation was offered by the Department of Theoretical Chemistry. The room, on the ground floor of 1 South Parks Road, has its own entrance from the outside; it has therefore been possible to provide a card-operated lock, offering 24-hour access to card-holders. After the necessary building work had been completed, the new Data Centre opened in January. The facility has proved quite popular, although there have been problems with the XBM connection to the lineprinter there. This will be replaced by a Cambridge Ring connection during the summer. The PAD - reverse PAD link used for the terminals has also experienced problems, notably with routing terminals through JANET to access remote sites.

ICL 2988

The most notable change to the 2988 service has been the introduction of the new terminal service at 4800 bps. This has greatly improved the usability of the system, and nearly all terminal access is now at this speed. Apart from the obvious advantages it opens up new facilities which are virtually unusable at low speed, such as browsing through HELP information. It is intended to transfer all terminal access to this route (users of terminals limited to 300 bps will benefit from enhanced terminal control facilities) but this change has been postponed until a solution is found for a software fault in the Camtec PAD that prevents output being interrupted on terminals operating at lower speeds.

Release 811 of VME was implemented in May. Although this release went fairly smoothly, there were found to be several software problems which caused unreliability during the summer. Delays in rectifying these have given rise to great dissatisfaction, and the Service has taken action to escalate procedures within ICL. At the time of this report, satisfactory responses to a number of faults are still awaited.

The delays in release of VME 811 impinged on development of the CAFS service. Two facilities were awaited; Direct CAFS Interface (DCI) and CAFS Search Option (CSO). The former provides a programming interface, allowing simple yet powerful programs to be written in any high-level language, using the full facilities of CAFS. A pre-release version of this was tested, and a fully supported version became available with VME 811. This has proved very successful. CSO is intended to allow existing programs and packages to be run in an environment where their data are pre-selected for them, using commands within the SCL to drive the CAFS unit. This would have enormous benefits for many users of statistical packages. Unfortunately when CSO was released it was found (despite earlier assurances from ICL) that it was only usable in a Cobol environment. This matter is being pursued with ICL.

A report on the use of CAFS is included as Appendix B.

Three new compilers - APL, Prolog and Lispkit - were mounted on the 2988 during the year. New versions of TSP, MDS(X), GENSTAT, SIR, GIMMS, CLUSTAN, ASPEX and the NAG FORTRAN library were also introduced.

VAX

A Systems Industries 414 Ms Winchester disc drive was added to each of VAX1 and VAX2 in September. Making use of these involved substantial changes in the layout of system files and user filestore, and also a revised back-up strategy. Once these changes had been carried out, the additional filestore came into service in November.

In March VAX1 and VAX2 were upgraded to 11/785 processors. The 11/785 is, for average user work, some 30-40% faster than an 11/780. It was decided to make no change to the resource allocation scheme or to the resources allocated, so that the increased power would be available to existing users in the same proportion as their previous resources. However, the job limits on the batch queues were reduced, from six hours to five hours for the Weekend queue, and from three hours to two and a half hours for the Large and Free queues. These limits were chosen so that all existing jobs would continue to run. No change was made to the Medium and Small queues. In the event the changes had to be delayed because of the performance deficiency in the newly installed processors, which was rectified in April. Since then the 11/785s have performed well. The accounting figures show a decrease in interactive mill-time, which seems to indicate that the previous workload is being done in less time. Batch work, as always, expands to fill all the remaining time available.

As part of the same upgrade a Floating Point Accelerator was added to VAX3, the remaining 11/780. The memory on VAX1 was converted to the newer 64K chip memory (this upgrade had been carried out previously on the others) and all three VAXs had their memory increased to 8 MB. In May a 16-line Able VMZ32 buffered terminal controller was added to each VAX.

Version 4 of VMS arrived in January, and a large amount of work preparing for its introduction was done in dedicated sessions on VAX3. This enabled VAX3 to be converted to VMS 4 as soon as the summer term was ended. Having this machine running the new version of VMS all the time allowed much more testing to be done, especially of the communications software. It is planned to convert the main service machines at the beginning of September.

VMS 4 contains many minor improvements to the user image and some major underlying changes, particularly in the area of security.

The Digital C compiler, the algebra package REDUCE and the simulation package SIMAN were mounted for the first time. New releases of SAS, TSP, P-STAT and the NAG FORTRAN and Algol68 libraries were introduced.

[Top]

User Liaison

Documentation

New software and new services of other kinds are added every year. With new services comes the need for new documentation; the number of User Guides published by the Service is now almost 70. Those currently in print are listed in Appendix C.

Remote sites

Allocation of time at ULCC and UMRCC continued to exercise Service staff and University committees. For the 1984-5 allocation year the University bid for, and received, 4.5% of ULCC (taken almost all as Cray time) and 4% of UMRCC. During the year extra time was secured at both National Centres, earmarked for specific SERC grant holders. In the case of UMRCC, the project was a very large calculation on the Cyber 205 by a research group in Theoretical Physics. Despite the exploitation of the Cyber 205 for a small number of problems of this kind, the resource in most demand is consistently the Cray at ULCC. After careful evaluation of proposals from departments, the University decided to bid for 6% of ULCC and 4% of UMRCC for the 1985-6 allocation year. UMRCC were again able to award the 4% requested, but at ULCC the Cray was oversubscribed by about 20% in total and the University has provisionally been awarded only 5.1% of the centre’s resources.

As noted elsewhere in this report, job transfer to both national centres is now effected by way of the Scicon HASP board in VAX3. The advent of HASP access to the Amdahl front end at UMRCC meant that users could be moved onto the Amdahl at an early stage in the rundown of the old 1904S front-end machine. This move was carried out in August and September 1984. Users then had ample time to become accustomed to the Amdahl service and to transfer their 1904S files before the 1904S was withdrawn at the end of April. When the ULCC service was transferred to the same HASP link in December, there was an initial period while the link was being switched manually between the two sites, and a timetable suiting both users of London and of Manchester proved tricky to establish. The development of software and hardware to switch automatically between the services according to demand has successfully solved this problem, and the HASP service as a whole has now been stable for some time.

Courses and Seminars

During the year the old ‘dumb’ terminals in the Computing Service lecture room were replaced by BBC microcomputers equipped with the Uniterm emulator chip. Uniterm (developed at Swansea) offers the choice of DEC VT100 or Tektronix 4014 emulation, so that the room can now be used for courses involving screen mode software on the VAX (such as the EDT editor) or graphical software.

The staple introductory courses - Getting Started with ECCE, Getting Started on the 2988 and Getting Started on the VAX - were again given throughout the year. A more advanced course, entitled Further Use of the 2988, was introduced for the first time. It was given three times, in Hilary and Trinity terms. Preparatory work has been carried for another new course, Getting Started with EDT.

Various lectures and seminars, at various levels, were also arranged. In chronological order through the year, they were:

OctoberSPSS to SPSS-X conversion course
NovemberOpen presentation for new users of the Computing Service
MarchSeminars on database topics (database design, types of database management system, PDS, IDMS, QUERYMASTER, FAMULUS, SIR)
May-JuneSeminars on special topics (array processing with the MAP-6420, the CAFS DCI programming interface, the SCL command language for VME, and library automation)

The SCL seminar will in future be run regularly in conjunction with the course Further Use of the 2988.

[Top]

Computing in the Arts

Lasercomp

The workload on the Lasercomp continues to grow and several large volumes from it have been published. Copies of all published work are kept at the Computing Service.

The new version of LASERCHECK was installed in the spring and since then both it and the old version have been run to enable users who had started on the old program to finish their work. All new characters are now software-sized and there have been some additions to the OUCS font library. Input to the slave software is now possible via JANET.

External users’ meetings have been held twice and the following Lasercomp documentation produced: two issues of Lasercomp News, amendments to the LASERCHECK Users Manual for NEWLASERCHECK, a document summarising the changes in the new software, Access to the Lasercomp via JANET, an introductory leaflet on the Lasercomp service, and a revised set of Lasercomp forms. Two Lasercomp courses were given during the year.

Work on non-Latin alphabets has continued. Syriac is now available but some characters are still missing for Devanagari. A paper on ‘Typesetting Exotic Scripts at Oxford’ was presented at a conference in Dublin in October.

A conference on Typography and Design Methods for Computer Typesetting was organised by the Service in April as part of the IUSC Workshop series. Just over 90 people attended. Staff have participated in the BCS Electronic Publishing Group and in a British Academy Working Party which has prepared a booklet on ‘Word Processing and Publishing: Some Guidelines for Authors’. A paper on the Lasercomp service was given at the International Conference on Computing in the Humanities at Brigham Young University, Utah, in June 1985.

Kurzweil Data Entry Machine

The purchase of the new processor referred to in last year’s report was approved by the Computer Board, and installation took place in October. Another addition during the year was a tablet which enables sections of a page to be selected for scanning, thus removing the need to mask unwanted areas or to cut the page; the tablet is particularly useful for dealing with columnar material or publications with footnotes. The new processor has improved the scanning rate considerably, and the hourly charges have been adjusted to take account of this. However, the overall throughput was reduced because of staffing difficulties. The decision has now been taken to augment the full-time staff of the unit to four.

During the year 44 jobs were scanned. The largest of these were the Decretum Gratianum and a project involving the scanning of numerous articles from the New Scientist, taking 236 and 309 hours respectively.

Oxford Concordance Program

The program was issued to a further 31 sites; 5 in the UK, 12 in the United States, 4 in Australia, 2 in South Africa and one each in Austria, Finland, Germany, Japan, the Netherlands, New Zealand, Norway and Thailand. There were 7 re-issues to sites in the UK. All OCP issues are now recorded in a PDS database.

In preparation for Version 2 and the projected microcomputer version, a questionnaire was sent to all then current sites in the autumn. A response of about 50% yielded a number of suggestions for the new version

Oxford Text Archive

The Shortlist was updated twice and the pricing structure revised to reflect the amount of work required when several texts are ordered by one person. In all, 65 new texts were added to the Archive and 239 texts were issued. Of these 48 new texts came from abroad and 133 issues were to scholars outside the UK.

Some reorganisation of the Archive tapes has been carried out.

Courses and Lectures

The SPITBOL course was given twice, once each in Michaelmas and Trinity terms, at the Computing Teaching Centre. The courses Text Analysis and the Computer and Computer Methods for Historical Research were given once each, as was a special version of the Text Analysis course for first year graduate students in Modern Languages.

Besides the activities noted above under the Lasercomp heading, members of staff have lectured and given seminars at a number of universities and other institutions, both in this country and abroad. Among these were an invited paper at the conference held to celebrate the twentieth anniversary of the Literary and Linguistic Computing Centre in Cambridge (1-2 October 1984), and an OCP workshop immediately preceding the International Conference on Computing in the Humanities (Provo, Utah, 25 June 1985).

[Top]

Performance and Reliability

2988 Service

Hardware reliability has in general been much improved after the difficulties of the previous year. One FDS640 head-disc assembly was replaced, but this was accomplished with no loss of user filestore. Several intermittent and obscure software problems which have followed the introduction of VME release 811 are a source of current concern, and ICL are being pressed for an early resolution of these problems.

VAX Service

The head-disc assemblies on the 512 MB Winchester discs of both VAX1 and VAX2 have been replaced during the course of the year. Apart from these incidents, the main interruptions to the user service have been associated with the installation of the various hardware enhancements. Some of these, notably the VMZ32 buffered terminal controllers, gave rise to periods of minor faults which were cleared relatively quickly.

Other Equipment

Communications equipment and graphical devices have for the most part performed reliably. Two of the older items, the Modular One communications processor and the Versamat film unit used in microfiche production, were removed from service with the Versamat being replaced by a new Cordell unit. These changes have eliminated two of the major maintenance headaches of recent years. The only problem worthy of note has been the continued unreliability of the Lasercomp’s magnetic tape deck; this fault is the more aggravating, for both users and Service staff, because of the poor response of the Monotype service organisation.

Summaries of machine usage for the year are given in Appendix D.

[Top]

Appendix A. The ICL 2988 Replacement Working Party

Dr M.B. Powell (Chairman) - Mathematical Institute
Professor J.M. Brady - Engineering Science
Dr R.J. Cashmore - Nuclear Physics
Professor D.F. Hendry - Institute of Economics and Statistics
Dr R.W. Hiorns - Biomathematics
Professor C.A.R. Hoare - Programming Research Group
Mr A. Jones - Oriental Institute
Dr C.K. McPherson - Community Medicine
Dr P.A. Madden - Physical Chemistry
Professor K.W. Morton - Numerical Analysis
Mr S.D. Page - Faculty of Music
Dr C.M. Perrins - Zoology
Dr C.E. Phelps (Secretary) - Computing Service
Dr A.G. Robiette - Computing Service
Dr A.J. Skarnulis - Chemical Crystallography
Dr G.L. Taylor - Molecular Biophysics

[Top]

Appendix B. A Report on CAFS Usage at Oxford.

To date, applications have fallen into three categories:

IDMSX Databases

Lexicon of Greek Personal Names Project (IDMSX database, 120 MB)

This is a major research project of the British Academy, the aim of which is to produce a five-volume record of all Greek personal names attested from earliest historical times up to the mid 7th century A.D. Typesetting of the work on the OUCS Lasercomp phototypesetter will be database driven. It is also intended that the database will offer a shared, online enquiry service to other interested scholars. QUERYMASTER and CAFS have two major roles to play in this project:

Much of the design and implementation of this database predated the arrival of CAFS at OUCS. Nevertheless CAFS/QUERYMASTER has been a valuable add-on extra at little or no programming cost, particularly as a means of producing reports which would otherwise have required special programming.

Beazley Archive Database Project (IDMSX database, 50 MB)

This project, which is funded by the Getty foundation, aims to computerise the catalogue of Greek vases begun by Sir John Beazley in the 1950s. The catalogue enables scholars to trace the whereabouts and cataloguing histories of all surviving vases, indexed by their iconography and the artists to whom they have been attributed. As with the Greek Lexicon, the project has been in progress for several years, initially using the simple flatfile-based information retrieval system Famulus.

The facilities offered are also very similar to those of the Greek Lexicon project: the database is updated by a simple interactive program using FORTRAN DML and interrogated by QUERYMASTER, which is also used to produce lists of hit records for input to a specialised report program. One novel feature is the use of CAFS text searching capabilities: a full text description of the iconography of each vase is held in the database, so that the QUERYMASTER user can request information about (say) vases decorated with pictures of Herakles and the lion: controlled keyword indexing is also used to ensure consistency.

The database was demonstrated (live) at a recent international conference of art historians in Pisa, where it reportedly aroused considerable interest.

The Ashmolean Project (prototype under development)

Following on the success of the above two projects, a design has recently been initiated for a full scale IDMSX/CAFS database system to support all aspects of the cataloguing and conservation records of the Ashmolean Museum. This is the first IDMSX database which has been designed specifically with CAFS in mind, and it will exploit such features as text quite heavily; this is seen as particularly appropriate for this application which is typified by partial and variable data.

Implementation of the pilot project, restricted to the Museum’s ancient Egyptian collections, began in summer 1985 using Application Master, DCI and QUERYMASTER.

PDS/QUERYMASTER and ISAM Files

The design and maintenance overheads of full-scale IDMSX systems would not be appropriate for many of the research applications supported, most of which are single-user systems of limited duration. For such users, particularly those unfamiliar with data processing, PDS has been found to be very suitable entry-level software. In many cases, however, the large quantities of data involved can lead to serious performance problems on the heavily loaded 2988. In the absence of any intention on the part of ICL to provide CAFS searching directly from PDS, several of the larger PDS users have been provided with query views, enabling them to access their PDS databases via QUERYMASTER and CAFS.

This has proved to be a particularly worthwhile exercise. PDS can continue to be used for database updates whilst QUERYMASTER is used for fast interactive retrievals. These can also take advantage of the improved fuzzy matching capabilities and of the output record-type facility, which are supported by QUERYMASTER but not by PDS110.

The exercise has highlighted some irritating differences between the two basically very similar interfaces. An ICL development which is urgently needed is a packaged PDS to QUERYMASTER bridge, and it is also essential that the retrieval facilities of the two products should stay in step.

Some typical applications which have benefited from the PDS/CAFS/QUERYMASTER combination are briefly described below.

The Iowa Database (5 tables, 2.5 MB)

This user is investigating the growth of Republicanism in 19th century Iowa. The database contains information on 70,000 votes cast in the Iowa legislature over a period of 30 years. Multi-dimensional scaling and other techniques are used to analyse the voting patterns of representatives from different social and political backgrounds on a variety of issues. The project was the subject of a paper presented at the recent International Conference on Databases in the Humanities and Social Sciences in the U.S.

Old English Glossaries (3 tables, 4.5 MB)

This user is working on a large collection of Mediaeval manuscripts in which Latin and Old English words are glossed in various different ways. QUERYMASTER is used to select groups of words which have been glossed in a particular way, or to identify the different ways in which a particular word has been glossed. The information .provides important evidence in historical linguistics

Social Mobility Database (12 tables, 8 MB)

This user is attempting to quantify various assertions about social mobility and class divisions over a twenty-year period in 19th century Philadelphia. Data about membership in various elites and ownership of property and businesses was taken from such sources as street directories and social registers and entered in a normalised form. Simple time-series analyses are performed to demonstrate the changing social structure of particular wards. The unpredictable nature and complexity of the queries involved have justified our faith in both the flexibility of the PDS/QUERYMASTER relational view of data and the performance benefits of CAFS.

In-house Databases (B tables, 1 ISAM file, 10 MB; 7 tables, 1 MB)

OUCS offers a number of facilities in support of computer aided research in the humanities. Administration of two of these, the Text Archive and the Lasercomp typesetting facility is also supported by PDS/QUERYMASTER/CAFS databases.

TOMES, the Text Archive database, is used to maintain information about the origins and usage of our extensive archive of machine readable texts; it was described in a paper presented at a recent international conference on the use of computers in literary studies. As well as the tables used to monitor issues and receipts of text copies and to produce the Archive’s printed catalogue, the TOMES query view also includes bibliographic data for some of the texts, which is held as CAFS-searchable text in a conventional ISAM file.

LASER, the Lasercomp database, is used to maintain all the information needed to run the national typesetting service efficiently. Monthly accounts and project registration and usage figures are all produced from the same system. In both cases, the CAFS/QUERYMASTER combination has proved invaluable as a means of providing ad hoc reports and enquiries at minimal cost.

DCI Systems

Major perceived weaknesses of QUERYMASTER in our environment are its poor support for textual data, its dependence on Data Dictionary and the comparative difficulty of integrating it with existing scientific software. Many of the users write their own programs, usually in FORTRAN or Algol68; others rely upon widely used scientific and statistical software packages such as TSP, SPSS or SIR. Clearly interfacing all this software to the QUERYMASTER view of data is not a task to be undertaken lightly, if at all.

By contrast, the DCI product enables any competent programmer to tap CAFS search power directly from any high level language, making only minimal changes to existing code.

The three projects described below demonstrate three of the DCI product’s attractive features: firstly, its performance advantages over even QUERYMASTER and CAFS for simple applications, secondly, its comparatively sophisticated range of text searching facilities, and thirdly, its flexibility.

The Futures Markets Database (4 tables, 4 MB)

This is used by an econometrician working with commodity market time-series data. The aim is to investigate the costs and benefits of different pricing strategies in international markets. PDS was originally used for database creation and management and, under this regime, extraction of one year’s data for a single market took, typically, 240 seconds of OCP time. Provision of a query view and CAFS access by QUERYMASTER reduced this requirement to around 60 seconds. Since the arrival of DCI all data extraction has been by a FORTRAN 77 subroutine called directly by the analysis package and this takes 5 seconds for the same extraction (involving 3 ISAM files). As a result of this, the user is now able to pursue various subsidiary analyses which would have been prohibitively expensive without CAFS.

Text Systems

Included in the Text Archive mentioned above are many major literary works, analysis of which has hitherto been performed only by using expensive indexing and concordance generation software. As part of the evaluation of DCI, a simple text retrieval package was written in Algol68 and made available for use with a variety of texts. This software and its applications are to be the subject of an article in the ICL Technical Journal.

The largest text database accessible in this way is the Shakespeare corpus, about 15 MB of running text now held as a single CAFS-searchable sequential file. Scholars working on the new critical edition of the canon have found this a particularly useful tool both in the quantification of existing hypotheses about the transmission of the text and in the formulation of new hypotheses.

Also accessible by the same software are two large library catalogues (6 and 8 MB sequential files) prepared in CAFS-searchable form to enable Oxford’s library community (now entering the throes of extensive computerisation) to assess for itself the potential of CAFS.

Performance of all these systems has proved very impressive, the response time at a MAC terminal being limited more by the number of hit records returned than the size of the file, provided this is less than 20 MB or so. It is our view that larger files will still require some degree of secondary indexing to achieve reasonable performance.

The CSO Macro-Economic Time-Series Data Bank

This data bank which is freely available to academic economists from the ESRC Data Archive at Essex University comprises around 11,000 sets of time-series data relating to all facets of the U.K. economy. The data are held in variable length records, making it difficult to access them from standard statistical software packages. The object of the project is to make easily available to a group of around 20 interested researchers datasets of particular importance to their own field of study. For example, ‘find all quarterly sampled datasets at 1980 base prices relating to domestic fixed capital formation’. Extracted data will then be passed to specialist time-series analysis software for further processing.

It is anticipated that this database will be 15-30 MB in total size and will be updated quarterly. The retrieval package will use DCI from FORTRAN77.

[Top]

Appendix C. Current User Guides

A1.1 Introduction to the Computing Service

A2.1 Rules of the Computing Service

A3.1 Service Facilities

A4.1 Glossary of 2988 Terms

A4.2 Glossary of VAX Terms

A5.1 2988 Budgeting, Accounting and Scheduling

A5.2 VAX Budgeting, Accounting and Scheduling

B1.1 Oxford Command Specifications

B2.1 Getting Started on the 2988

B2.2 Further Use of the 2988

B2.3 SCL

B3.1 Editing Files

C2.1 Running Algol68 Jobs

C3.1 Running FORTRAN Jobs

E1.1 PDS: Personal Data System

E2.1 BMDP Statistical Software

E2.5 MDS(X) - Multidimensional Scaling Programs

E2.8 CLUSTAN: Cluster Analysis Package

E2.9 SPSS-X

E3.2 SYMAP

E3.3 SYMVU

E3.4 GIMMS - General Purpose Geographic Processing System

F1.1 Using Remote Computers from the 2988

F3.1 Using the PAD

G1.1 Creating and Editing Files using ECCE

G1.2 Getting Started with ECCE

G2.1 VAX/2988 File and Job transfer

G2.2 Listing Files to Film

G2.3 The Oxford File Archive

G2.4 Magnetic Tape on the 2988 and VAX

G3.1 Spitbol

G4.1 OCP - Oxford Concordance Program

G4.2 QUILL

G4.3 P-STAT - Princeton Statistical Program

G4.4 GENSTAT - A General Statistical Program

G4.5 TSP - Time Series Processor

G4.6 GLIM - Generalised Linear Interactive Modelling

G5.1 NAG Library

G5.2 GHOST80 Graphics at OUCS

G5.3 GINO-F Graphics at OUCS

M0.3 SPSS to SPSS-X Program Conversion

M1.1 OUCS 2988 Algol68 Library

M2.3 GPCP

M2.4 FACSIMILE

M2.5 MACSYMA: An Algebraic Manipulation Package

M2.6 REDUCE

M3.1 Calcomp-on-GHOST80 on the VAX and 2988

M4.2 Standards for Magnetic Tape transfer

M4.4 Lasercomp Work submitted from Remote Sites

M4.5 Facilities for Reading non-VAX Tapes on the VAX

M5.2 Trident TT100 Brown Terminals

M5.3 The Qume Printer

M5.5 Dacoll M249 Graphics Terminals

M5.6 The Calcomp 1012 Plotter

M5.8 Tektronix Graphics Terminals

M5.10 Datatype X5A Colour Graphics Terminals

M5.11 BBC Micro Terminals

M5.12 Xerox (EPS) 2700 Laser Printer

M5.13 Hewlett Packard 7475 Plotter

M6.2 Magnetic Media Conversion Service

M6.3 File Transfer between an Apple and a VAX, using Kermit

M6.4 Cifer File transfer for CP/M Discs to the VAX

M6.5 File Transfer Using the Cambridge Ring (reading CP/M discs on 380Z)

M6.6 File Transfer between an IBM PC and a VAX using Kermit

M7.1 Cards and Paper Tape

V1.1 Getting Started on the VAX

V2.1 Using the VAX

V3.1 SAS: Statistical Analysis System

V3.2 MINITAB - An Interactive Statistics Package

[Top]

Table 1. Machine Usage

4-Weeks Ending

2988

VAX

Interactive

Batch

Interactive

Batch

 

No. of Jobs

CPU Hours

No. of Jobs

CPU Hours

No. of Jobs

CPU Hours

No. of Jobs

CPU Hours

29/7/84

13622

60.75

11678

461.62

17530

254.64

6047

691.15

26/8/84

12213

67.09

10704

503.22

14790

217.00

4646

450.27

23/9/84

11526

73.76

9569

409.50

13856

193.06

4023

509.91

21/10/84

12541

73.56

10293

557.06

15751

206.61

4110

512.12

18/11/84

14305

82.56

10011

594.10

21788

269.58

5367

750.43

16/12/84

14675

97.08

9943

510.76

25208

293.18

5522

698.45

13/1/85

7363

55.07

5660

273.60

12557

152.21

4480

641.01

10/2/85

15461

112.39

10333

592.03

23508

291.08

6847

841.52

10/3/85

15853

126.69

11361

512.04

23818

308.10

7959

858.13

7/4/85

13729

112.65

10892

588.37

20797

244.14

7043

871.53

5/5/85

12077

80.16

9766

560.71

20804

217.26

7773

910.51

2/6/85

10918

78.29

8570

420.72

23907

229.02

7263

905.28

30/6/85

12651

99.99

10517

541.39

23329

242.86

7295

860.05

[Top]

Table 2. Reliability and Mean Time Between Failures

Reliability is the percentage of scheduled productive time (i.e. switched-on time not allocated to maintenance, development or backup) during which the machine is fault-free. Weighting factors are applied to equipment failures that allow the service to continue with impaired performance. Mean Time Between Failures (MTBF) is the mean time between incidents requiring a system reload to take place.

4-Weeks

2988

VAX1

VAX2

VAX3

Ending

Reliability (%)

MTBF (hrs)

Reliability (%)

MTBF (hrs)

Reliability (%)

MTBF (hrs)

Reliability (%)

MTBF (hrs)

29/7/84

91.36

61.61

97.92

168.00

97.96

335.50

95.21

106.90

26/8/84

95.05

61.68

99.56

335.35

95.07

134.40

99.54

89.33

23/9/84

99.65

117.03

99.96

336.00

94.10

95.39

99.94

536.00

21/10/84

97.44

82.34

99.85

224.00

96.41

74.67

99.50

58.00

18/11/84

99.81

168.12

98.73

224.00

98.13

224.00

97.83

336.00

16/12/84

99.17

122.79

99.98

672.00

100.00

¥

99.94

336.00

13/1/85

95.69

53.06

97.69

95.75

98.01

168.00

98.02

112.00

10/2/85

99.09

159.22

99.00

168.00

99.86

134.40

97.69

111.58

10/3/85

94.79

68.45

98.72

168.00

98.66

96.00

97.94

96.00

7/4/85

99.08

116.39

95.13

335.50

95.42

224.00

93.07

96.00

5/5/85

98.98

151.40

99.69

336.00

99.98

335.00

95.18

41.87

2/6/85

93.69

53.55

98.40

96.00

97.32

96.00

96.53

84.00

30/6/85

99.00

119.17

88.32

672.00

99.92

672.00

96.02

39.53

[Top]

Table 3. ULCC Usage

4-Weeks

Amdahl

Cray

Ending

Jobs

%Jobs

Units

%Units

Jobs

%Jobs

Units

%Units

29/7/84

555

1.0

295.2

0.3

214

1.5

*

3.0

26/8/84

477

0.9

290.4

0.2

182

1.3

1030.7

0.2

23/9/84

672

1.4

407.0

0.3

262

2.2

10607.6

2.9

21/10/84

743

1.4

448.9

0.5

290

2.1

11453.9

3.3

18/11/84

1028

1.7

655.5

0.7

433

2.7

36305.2

10.3

16/12/84

932

1.5

416.4

0.4

475

3.1

48459.4

12.8

13/1/85

266

0.7

57.5

0.0

113

1.5

7738.9

2.5

10/2/85

902

1.3

286.6

0.2

394

2.5

32392.3

7.9

10/3/85

1663

2.3

1316.2

0.9

306

1.7

8230.0

1.9

7/4/85

1679

2.5

1728.9

1.6

297

2.0

23324.9

6.1

5/5/85

1857

3.1

1547.7

1.6

445

3.0

20473.1

4.5

2/6/85

1198

1.9

615.0

0.7

517

3.4

51072.5

12.2

30/6/85

1446

2.2

1349.3

1.4

364

2.1

50503.5

11.2

* Cray accounting was carried out on a different basis prior to 1/8/84

[Top]

Table 4. UMRCC Usage

4-Weeks

CDC 7600

Cyber 176

Cyber 205

Ending

Jobs

%Jobs

Units

%Units

Jobs

%Jobs

Units

%Units

Jobs

%Jobs

Units

%Units

29/7/84

605

9.3

147131

29.8

*

*

*

*

94#

*

997#

*

26/8/84

760

10.3

223212

39.8

*

*

*

*

5#

*

20#

*

23/9/84

605

9.7

131382

27.0

*

*

*

*

8#

*

108#

*

21/10/84

*

*

*

*

*

*

*

*

*

*

*

*

18/11/84

845

2.1

202212

20.1

*

*

*

*

*

*

*

*

16/12/84

394

1.7

144153

25.2

24

0.1

936

0.1

83#

*

3639#

*

13/1/85

303

2.2

94085

31.3

25

0.1

1677

0.3

94#

*

7100#

*

10/2/85

560

2.4

185149

26.8

70

0.3

3497

0.5

*

*

*

*

10/3/85

331

1.3

110944

14.6

69

0.2

6499

0.8

*

*

*

*

7/4/85

310

1.4

49068

8.3

38

0.2

4463

0.7

81

3.3

21856

5.6

5/5/85

491

2.3

119235

18.7

99

0.4

9355

1.4

213

3.3

85774

6.6

2/6/85

430

2.1

132341

20.6

*

*

*

*

232

2.9

107997

7.2

30/6/85

362

1.5

107447

18.3

26

0.1

1450

0.2

238

2.3

189197

10.3

* Usage figures not available

# Statistics incomplete

[Top]