The EASE Information Directory and the EASE Enquiry Service are being set up as a direct result of requirements identified during visits to Higher Education Institutes. As they come on-stream, please make use of these facilities to ensure their continuity.
As a result of the Workstation Assessment activities, you will note that DEC, IBM and HP have been added to the Approved List.
We wish you a Happy Christmas and a successful 1991. Special thanks go to the many contributors throughout 1990. The New Year issue will include an updated index of all contributions since ECN was started in 1987.
Issue 24 of this Newsletter carried an Update and Questionnaire on the EASE Information Directory (EID). Its purpose was to:
The answers we received from Newsletter readers and from Higher Education Institutes (HEIs) contacts, who were sent questionnaires directly, made an important contribution to our Feasibility Study. We were very pleased in particular with the response to our crunch question; only one respondent said he would not anticipate using the service, and this was someone who does not have access to JANET.
The Working Party (which consisted of Subodh Chanda, John Kalmus, Bob Maybury and leader Debbie Thomas) completed its Feasibility Study at the end of August with a report to the Computing Facilities Technical Advisory Group, which in turn made recommendations to the Computing Facilities Committee. Copies of these papers will soon be available as RAL Reports. The main points can be summarised as follows:
A detailed workplan has been drawn up to take these proposals through into implementation. The hardware and software required have been identified and will be available before Christmas. A new project team, under the leadership of Dr Keith Jeffery, Head of the Systems Engineering Division of the Informatics Department at RAL, will be responsible for this new phase. We hope to provide a prototype User Interface for demonstration at EASE 91 in Birmingham next March.
The netlib software distribution system has now been running at the University of Kent for ten months. During this time it has processed approximately 8000 requests resulting in over 180MBytes being e-mailed to 830 different individual users. The software has been steadily expanded throughout the lifetime of the system. New software includes:
along with updates and additions to many of the existing libraries including the ACM Toms algorithms, matlab and benchmark sections. A file containing all the changes and updates to the systems may be requested from netlib.
In June the server was moved to a SUN 4 and the distribution software upgraded at the same time. This change was transparent to users and was effected without interruption to the service. The new software both improves the way in which large requests are split up prior to e-mailing and allows us to offer automatic program execution facilities. The first program to be made available in this way is the AT&T Fortran to C translator, this allows users to e-mail netlib Fortran source and receive back the resultant C. We intend to extend this facility in the near future to include a number of Fortran tools for formatting Fortran source, transforming single precision code to double precision and checking code for adherence to the ANSI standard.
A full list of the available libraries may be obtained by sending a mail message to netlib@uk.ac.ukc whose body contains the single line
send index
The file containing details of new and updated software can be retrieved via the message
send updates from updates
To use netlib's execution service send the following message
execute f2c
and immediately follow it with your data for f2c. Thus a complete mail message might be:
execute f2c subroutine test integer x x=1 end
More details of the execute facility may be obtained using
send index from exec
This file will be updated as more facilities are made available.
We would very much like to know how members of the EASE community view this system. Does netlib provide a useful service to the community? Is there software which you would like to obtain which isn't available from netlib? Are there other facilities you would like to see provided by netlib?
We should also like to set up a mailing list of EASE netlib users - so if you have used, or intend to use, netlib as part of any EASE related project please would you let us know.
Answers to these and many other questions can be directed towards the EASE Enquiry Service from Wednesday, 2 January 1991.
This new facility will attempt to provide the EASE Community with answers to questions on all aspects of the EASE Programme, its services and other activities, either by telephone or bye-mail. Where answers cannot be provided immediately details will be passed onto a local expert and hopefully a response obtained reasonably quickly.
This new facility, as are all the EASE supported activities, is available to anyone in the Engineering or Computer Science Departments at HEIs, who either uses or intends to use computers/workstations/etc in their research or other activities. It is not necessary to have an SERC research grant to use this service.
The telephone line will be manned from lOam to 4.3Opm Monday to Friday with an Ansafone facility for out-of-hours enquiries.
The EASE Schedules were first defined three years ago and represented a cornerstone of the original EASE programme. They contained the hardware and software approved and supported by the Computing Facilities Committee (CFC). Significant changes have taken place over the last three years, both in the computing resources available to engineers and in the approach adopted by CFC to the provision of these resources. It was therefore decided earlier this year that a major revision of the EASE Schedules should take place to reflect these changes. This process is now nearing completion. As you read this, the committees and sub-committees of Engineering Board will be reviewing a provisional release of version 2 of the Schedules with a view to a full release to the community early in 1991.
A significant trend in recent years has been the move away from a prescriptive approach to the supply of resources towards one of providing as much help as possible to the researcher to enable him to make a sound choice for himself. The new Schedules reflect this by providing information and advice on a wide range of hardware and software of potential interest to engineering researchers. The results of the continuing EASE hardware assessment programme are reflected in recommendations under a number of categories of equipment including workstations, servers, X terminals and peripherals. The experience within the community and at RAL in using a wide range of software from language compilers, through software development tools, utilities and libraries to applications packages is reflected in a comprehensive range of software entries. The information supplied in each entry is also considerably expanded to include details of how to get the software and support for it, an indication of the cost, the hardware on which it is available and other items. It is intended that the range of applications packages included should cover the needs of all sectors of the Engineering Board programme.
It is strongly recommended that prospective grant applicants should consult the Schedules before specifying their hardware and software requirements. In future, the Schedules will be regularly updated with the assistance and advice of the Computing Facilities Technical Advisory Group (CFTAG), so that they continue to reflect current experience and respond to current needs. When the EASE Information Directory Service is available, the Schedules will be accessible on-line.
As noted earlier, the revised Schedules are at present under review. Anyone wishing to contribute to this process is invited to contact me. Copies of the provisional release of version 2 can be made available with a response form for comments.
I have read the RAL server report, RAL-90-032, which has also appeared in this newsletter in Issue 23 (June 1990), with interest. However I am beginning to wonder whether the investment in central servers is really necessary, even in a large department.
I am seriously considering replacing all our SMD drives and servers with low cost alternatives.
Principal candidate would be:
Sun 4/65 processor box with 2-4 third party 1.2GB sync scsi. (Also exabyte, 2nd ethernet interface, 100MB internal drive sync scsi - slow but Sun provided and maintained, not used for any real purpose - plus some Sun approved media distribution device.)
This was not available at the time of your report. Indeed, Sun still do not list a 4/65S except bundled with a QIC-150 and 669MB sync scsi. However, I wonder if you have subsequently been able to test a sync scsi based configuration. If my understanding is correct, sync scsi devices should give a noticeable performance improvement over async devices, eg over the 327MB your report mentioned on the 4/370.
Have you any experiences that would refute or support this claim?
Thanks for your note. We are certainly interested in the ideas you put forward. Our only experience of SMD versus scsi was that reported in the Server Evaluation paper. We ran the tests on a Sun 4/370 initially with three 327 scsi discs, and then repeated them with one 688 SMD instead. The results were very similar, leading us to the conclusion that the multiple drives compensated for the slower interface.
It is very difficult for us to test third-party peripherals without buying them.
We do not plan to test a 4/65 in the configuration you suggest but we are certainly prepared to let you have our tests if you are in a position to carry out some testing. If possible, we would like to include those results in our paper.
This report is the latest in our series of Workstation assessments undertaken for the Computing Facilities Technical Advisory Group (CFTAG). Because of the size of the market, it was agreed that this particular exercise would only cover new machines. No attempt would be made to test systems which have not altered since the last exercise, and new models in an already tested machine range were also excluded. Thus this report does NOT provide a complete summary of the market. The exercise was planned in the same way: the completion of a questionnaire by the supplier, meetings between supplier and RAL, and benchmarking a loaned machine at RAL.
The following is a brief overview of the assessment. Much more information is contained in the full report, available on request.
It was decided to retain the Sun 3/60 with 8Mb of memory as base machine, for consistency with previous reports. However, this model is no longer supplied by Sun, and is underpowered by today's standards. In all the tables, the results for the base machine are supplied first, followed by the others, generally in alphabetical order of supplier. However, three of the machines tested were from the top of their respective ranges, and it has therefore been decided to present their results in a separate section of each table. Benchmark figures have again been presented as 'Which Report' style tables, normalised to the corresponding result for a Sun 3/60. The key refers to all the following tables. (NB Changes have been made to the top of the scale to allow for the increase in power of the workstations tested.)
| Code | Range of ratios to Sun 3/60 |
|---|---|
| Blank | 0.9-1.2 |
| + | 1.2-1.5 |
| ++ | 1.5-2.5 |
| +++ | 2.5-4.0 |
| ++++ | 4.0-6.0 |
| +++++ | 6.0-10.0 |
| ##### | >10 |
| * | No meaningful result |
| - | 0.7-0.9 |
| -- | 0.5-0.7 |
| --- | 0.3-0.5 |
| ---- | <0.3 |
Since the last report, Sun have produced the SPARCstation 1+, the SLC and the IPC. DEC have produced the DECstation 2100. Since these machines are part of ranges already approved, they were not tested this time. However, DEC have also produced a new range of machines, the DEC 5000 series, which was considered sufficiently different to be included in the assessment.
All the previous suppliers were contacted, as well as a number of new ones. Those who responded positively are listed below. We are grateful for their cooperation in providing machines for test. Intergraph, MIPS, Mistral and Tektronix did not have a suitable machine at this time.
| Supplier | Model | CPU | Clock KHz |
Store Mb |
Monitor (inches) |
Disc Mb |
Price Bracket |
|---|---|---|---|---|---|---|---|
| Sun | 3/60 | M68020 | 20 | 8 | 19M | 141 | * |
| Acorn | R260 | ARM3 | 30 | 12 | 19C | 100 | A |
| Data General | A Viion/310 | M88000 | 20 | 20 | 20M | 322 | D |
| HP | 9000/345 | MC68030 | 50 | 12 | 14C | 2*304 | C |
| HP | 9000/345 | RISC (HP) | 15 | 8 | Cifer* | 304 | E |
| Silicon | Iris 4D/25 | R3000 | 20 | 12 | 19C | 327 | D |
| DEC | Decstation 5000/200CX | R3000 | 25 | 24 | 16C | 665 | F |
| IBM | RS6000/320 | RISC (IBM) | 20 | 32 | 19C | 2*320 | F |
| Solbourne | 5/501 | SPARC | 33 | 32 | 19C | 661 | F |
* = HP were unable to supply a fully configured 9000/834 with monitor.
The Sun3/60 is no longer sold, so no price is available. A similar configuration for a Sun3/80 (which is more powerful) would fit into price bracket B.
The prices used are for the configurations of the benchmark machines, and do not include any form of discount. All provided C, Fortran and NFS. Acorn does not yet have hardware floating point (due in 1991).
For preference, all benchmarked machines should have similar configurations. However, it is not always possible for suppliers to provide the machine we would most like to have. This should be borne in mind when comparing results. The price brackets range from <£7K (A) to >£30K (F).
Tests were performed on the basic machine speed, performance of Fortran, Pascal and Ada compilers, GKS, X-windows and NFS. There is no space to display all the results, but, as an example, Table 2 gives relative values for basic operations from Fortran, including compilation speed.
The Acorn Fortran optimiser nullified the Function Call test. It also affected the Integer tests. Data General failed to complete the compile speed test. DEC completed, but the resulting program failed to run. The HP834 was affected by lack of memory.
DEC, HP and IBM have a good range of Engineering software available on their machines. Silicon Graphics is well-represented in some areas such as Mechanical Engineering.
Table 3 gives relative values for the performance of real applications. As well as the four tests used in previous exercises, it has been possible to run several additional examples. The tests are:
Therm would not run on either the Data General or the DEC.
| Machine | Fortran compiler |
Integer | FP | Function calls |
I/O |
|---|---|---|---|---|---|
| Acorn | +++ | +++ | -- | * | --- |
| Data General | * | +++ | ##### | +++++ | ++ |
| HP345 | ++++ | +++ | ##### | +++ | ++ |
| HP834 | +++++ | ++ | +++++ | ++ | +++ |
| Silicon Graphics | +++++ | ##### | +++ | +++ | |
| DEC | * | +++ | ##### | +++++ | +++ |
| IBM | +++++ | ++ | ##### | ++++ | ++++ |
| Solbourne | +++++ | +++ | ##### | ++++ | ++++ |
| Machine | Therm | FE | Image | IKBS | ERIL | LES | DFT | Upset |
|---|---|---|---|---|---|---|---|---|
| Acorn | ---- | --- | --- | --- | --- | --- | ||
| Data General | + | +++++ | +++++ | +++ | ++++ | --- | +++++ | ++++ |
| HP345 | ++ | +++ | ++ | ++ | ++ | +++ | +++ | ++ |
| HP834 | ++++ | ++++ | +++ | ++ | +++ | ++++ | ++++ | ++++ |
| Silicon Graphics | +++ | +++++ | ++++ | +++ | +++ | +++++ | ++++ | +++++ |
| DEC | * | +++++ | +++++ | ++++ | ++++ | ##### | ##### | ##### |
| IBM | ##### | +++++ | +++++ | ++++ | ++++ | ##### | ##### | ++++ |
| Solbourbe | +++++ | ++++ | ++++ | +++ | +++ | +++++ | +++++ | ##### |
Acorn is using software floating point.
The IBM results appear to be influenced by the ability of the data cache to hold the active variables and arrays. The Upset tests had to be run unoptimised (the optimised version failed).
The poor performance of Data General in the Linear Equations test is unexplained.
The HP834 results were good considering the small memory size.
Table 4 is an attempt to summarise the general characteristics and performance of each machine. This has been done by assessing each characteristic for each machine on a scale of * (poor) to ***** (excellent). A - indicates an apparent lack of the feature. The assessments are non-competitive, and do not represent a ranking.
The previous entry for the Sun 3/60 has been omitted to allow more scope for an adequate summary. Several of the machines are very much more powerful than the chosen baseline.
| Code | Title |
|---|---|
| 1 | Hardware |
| 2 | Software |
| 3 | Communications |
| 4 | Graphics Hardware |
| 5 | Graphics Software |
| 6 | X-Windows |
| 7 | Standard Languages |
| 8 | Additional Languages |
| 9 | Applications Available |
| 10 | Basic Systems Performance |
| 11 | Applications Performance |
| Machine | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Acorn | ** | *** | *** | *** | *** | *** | *** | *** | ** | ** | * |
| DG | **** | **** | *** | **** | **** | **** | **** | *** | ** | *** | **** |
| HP/345 | *** | **** | **** | *** | ***** | **** | ***** | ** | **** | *** | ** |
| HP | **** | **** | *** | *** | ***** | * | ***** | ** | **** | **** | *** |
| SG | **** | **** | **** | ***** | ***** | ** | ***** | ** | *** | **** | **** |
| DEC | ***** | **** | ***** | **** | ***** | ***** | ***** | ***** | ***** | ***** | ***** |
| IBM | ***** | **** | *** | **** | **** | ** | ***** | **** | **** | ***** | ***** |
| Solbourne | ***** | **** | ***** | *** | ***** | ***** | **** | *** | *** | **** | **** |
One of the major discoveries from this exercise was the emergence of a new range of workstations, such as exhibited by DEC, IBM and Solbourne, which seem to fit above the level exemplified by the Sun 3/60. The entry level price for such machines tends to be near the top of the price range for the other types, but the power increase is considerable. DEC certainly consider this to be a different type of system, since the DECstation 3100 is still the preferred general purpose desktop workstation. IBM on the other hand do not. With the appearance of cheaper Superworkstations, it is becoming less clear where the boundaries lie. Yet some form of selections seems necessary in order to keep the assessments within bounds. This subject will be discussed at the December CFTAG meeting. Suggestions from readers are welcome.
The original report was presented to CFTAG in September 1990. On the strength of all the evidence, recommendations were made to the Computing Facilities Committee (CFC) and these have been ratified. The DEC, IBM and both HP machines have been added to the Approved List. While recognising the capabilities of the other machines, the shortage of suitable Engineering software precludes their addition to the List at this time. CFC agreed that the Data General, Silicon Graphics and Solbourne machines would be reconsidered when their set of Engineering applications was more extensive, and that the Acorn should be retested when its hardware floating point is available.
A set of training materials and self-teach booklets have been developed by the Advisory Group on Computer Graphics. The master set includes all the documentation developed and gives permission for the documents to be copied for academic use. Alternatively copies of the booklets can be obtained. The materials include:
This is available as a master set of two box files for £50 for the UK academic community. The materials have been produced by people in the community and are especially designed for use in UK higher education.
EASE 91 is to be held at the University of Birmingham from 25 - 27 March 1991. Information on this event has appeared in previous issues of this Newsletter and regular updates will be provided in the future.
EASE 91 will be the 2nd Conference and Exhibition run under the EASE Programme and this year its theme is Development Environments for Engineering Applications.
EASE 91 starts on Monday, 25 March 1991 when there will be five Tutorial Sessions which relate to EASE-type activities. Tuesday morning, 26 March, will be given over to a Plenary Session with a number of Keynote speakers including Professor Mike Wozny of the Rensselaer Polytechnic Institute, NY. In the afternoon, the Conference will hear from the Teaching Company and from those who have benefited from the Teaching Company Scheme. The afternoon will close with a number of Exhibitors' Presentations.
There will be five parallel Seminar Sessions on Wednesday morning, 27 March covering a number of themes including Standards, Computing Environments and AI for Engineers. During Wednesday afternoon there will be meetings of the EASE Community Clubs and a Helpline. More infonnation on the Helpline is on this page.
The Conference will be brought to a close with a final keynote address from Dr B R Martin, Head of Engineering Division, SERC, who will talk about the Engineering Board's Future Directions, with relevance to IT.
Throughout Tuesday and Wednesday there will be an Exhibition which will include contributions from manufacturers, both of hardware and software, and from local Higher Education Institutes as well as a Book Exhibition. There will also be demonstrations given by manufacturers and demonstrations on the EASE support and development work being carried out at RAL.
Although EASE 91 is directed mainly at the academic community, we feel that it is also relevant to the industrial community. We would therefore seek to encourage attendance at EASE 91 from amongst those researchers in industry who would find the theme of the Conference together with the Exhibition of interest and relevance to them. As part of this, we would like SERC academic grant holders who work with industrial partners, to draw their attention to the Conference and Exhibition. There are no conference fees payable by members of the academic community, however, for non-academics there is a conference fee of £50 per day or £125 for the three days of the Conference.
EASE is setting about building a software development and run-time environment for use on approved workstations.
The first part will start by outlining the background and then focus on the impact it will have on user tasks. Areas of particular attention will be the process of developing new programs, giving a face lift to older ones and on the ways in which the environment will promote interworking between independent applications.
The second part of the tutorial will be given by very early users of the environment. These are the demonstrator projects which have been identified to see how the environment works on real practical problems. Talks will be given by project staff and users from HEIs.
As part of the EASE 91 Conference and Exhibition to be held at the University of Birmingham between 25 and 27 March 1991, two services will be available, which are designed to help the EASE Community with any computing software and hardware problems or difficulties that they may be experiencing in their research. The first is a Surgery Desk which will be open throughout the Conference and will be manned by RAL staff who will be available to discuss problems and provide advice to Conference participants on their computing needs. The other service, the Tools Helpline, open to all, will take place between 13:30 and 15:00 on the final afternoon of the Conference, Wednesday 27 March, running concurrently with the Community Club meetings. The Tools Helpline will cover those topics raised by Conference participants at the Surgery Desk where it is felt benefit would be derived from a more open discussion. Therefore those people who will be attending EASE 91, and who would like to take advantage of either or both of these two services, are invited to register now any computing problems or queries etc. that they would like to discuss at the Conference Surgery Desk or at the Tools Helpline. This will give an indication of the type of subjects to be raised and for arrangements to be made for the relevant RAL people to be present at the Conference.
A last minute reminder about this one day event which has been organised for 7th December 1990 at AIAI, Edinburgh. These one day events provide engineers with the opportunity not just to find out more about new products and techniques but also to meet and discuss common requirements with other researchers in their areas. This workshop is designed to highlight some of the newer techniques now being put into practical use within the engineering community. The topics covered will include genetic algorithms, neural nets and behavioural robotics. Talks will be presented at an introductory level, followed by discussion of their applications within the engineering disciplines. This event is free to academic engineers but numbers are limited.
We have developed a suite of Knowledge Based Systems courses that takes the student through the complete life cycle of knowledge engineering. The courses start at fundamental knowledge representation level, go through knowledge elicitation to the final knowledge engineering stage. The suite of courses consists of: Knowledge Based Systems Skills; Knowledge Elicitation; and Knowledge Engineering.
Additionally the language based courses have been enhanced. The suite of AI Language courses contains: Common Lisp; Introduction to Prolog; and Parlog.
The revised course schedule for 1991 is as follows:
Due to support from SERC via their EASE Programme some training places are available free of charge to academic engineers.
IEEE International 2-Day Workshop after Transputer Applications 91, 2-3 September 1991, Glasgow.
