Jump Over Left Menu
Issue 5: February 1997
NOTE: many of the urls referenced in this newsletter are no longer active.
The main feature in this edition is an introduction to the work of the European Research Consortium for Informatics and Mathematics (ERCIM), together with information on the ways in which we have participated in its activities in the past few years. ERCIM's members include some of the most prestigious IT institutes in Europe and the organisation has become a major force in the European IT scene, and we are privileged to be its UK member.
We include a brief summary of the most recent meeting of users of the Cray J90 (J90) service provided from the Rutherford Appleton Laboratory's Atlas Centre. Users should please note reference to a fuller report on the World Wide Web (WWW) in which the Chairman of the users group, Professor Catlow, encourages applicants to apply for the full allocations of resources their projects need on the J90, and observes that the costs of Cray vector processing time are substantially less on the J90 than they were on the Cray Y-MP. We also include in this edition articles on a couple of our EU-funded projects.
As this edition goes to press, we are hosting a meeting of the Executive board of the international World Wide Web Consortium (W3C), the body responsible for standardisation within the WWW. Bob Hopgood, who acts as CLRC's representative at W3C comments that "This will be the first Executive meeting held away from a Consortium site, and we're very pleased that W3C felt CLRC was the right organisation to host it,". A report will be included in our next issue.
As ever with ATLAS material, much more information on the topics can be found on the WWW pages quoted at the end of every article. Finally, and also on the subject of the WWW, the Department has re-worked its own WWW pages and these can now be accessed.
Brian Davies, DCI, CLRC
New DCI WWW page: http://www.dci.clrc.ac.uk. Cray J90 report from Prof Catlow: http://www.cis.rl.ac.uk/publications/ATLAS/feb97/catlow.html
News Feed Sites
The Department for Computation and Information (DCI) has recently been successful in a UKERNA tender to enhance the JANET Usenet Service by providing News Feed sites which will receive news feeds from commercial Internet Service Providers. DCI will operate News Feeds at Daresbury Laboratory (DL) and Rutherford Appleton Laboratory's (RAL) Atlas Centre with equipment supplied by UKERNA. These sites will feed news into and out of JANET to other selected organisations. The facility will be integrated into the existing operational support infrastructure which includes problem report tracking, single point of contact for client sites and management statistics. The service will be run by Penny Windebank at RAL and Rob Bradshaw at DL, both of whom have considerable experience in running News Feeds, with overall project management from Tim Kidd at RAL. Four other sites have also been chosen to run News Feeds. More details of the service to be provided by DCI will be given in a future issue of ATLAS.
Tim Pett, DCI, CLRC
CLRC and ERCIM
In 1989, three of the major Mathematics and Information Technology Laboratories in Europe (CWI in the Netherlands, GMD in Germany and INRIA in France) formed the European Research Consortium for Informatics and Mathematics (ERCIM) initially with the aim of coordinating their activities. From this grew a much larger grouping of European IT Research Laboratories to coordinate their research, to provide a conduit between academia and industry, to help Small to Medium Size of Enterprises (SMEs) and to provide a non-industrial and independent voice concerning priority areas in Europe. In 1990 Rutherford Appleton Laboratory (RAL) was invited to become the fourth member, and by the end of 1996 the membership has increased to the 14 institutes see diagram (the RAL membership is now vested in the Central Laboratory of the Research Councils (CLRC)).
What does ERCIM do?
In the initial phase, there was an emphasis on getting to know each other with a series of joint workshops on specific topics. Out of these has grown a set of permanent Working Groups which meet several times a year. The Graphics Working Group is the oldest of these followed closely by the ERCIM Database Research Group. They and some of the other Working Groups have received funds from the Human Capital and Mobility (HCM) Programme of the Commission for their activities. In the case of the Graphics Working Group this has allowed a coordinated response to ISO standards activities within Europe.
ERCIM Member Institutes
ERCIM has always been keen on making IT research a European activity. Very early on it launched its own Mobility Programme which gave Fellowships to young European researchers to spend time at three of the ERCIM Institutes working on a research programme that brought the three groups closer together. The HCM Programme provided additional funds which allowed the programme to be extended to a much larger number of researchers. This is quite a mature part of the ERCIM Programme now with the Fellows working closely with the Working Groups and on themes proposed by the Working Groups and approved by the Executive Committee.
To publicise its activities both within the Institutes and externally, an ERCIM Newsletter is produced four times a year and is available electronically to all interested parties. This gives information about the Institutes, their research programmes and also gives ERCIM opinions on European activities.
The closer links between the partners means that the number of joint research projects has increased and a number of initiatives have been instigated by ERCIM itself. One example is that ERCIM is the focus for a Digital Library initiative that relates to and works with the US initiatives in this area.
The European Commission has found ERCIM to be a useful vehicle to coordinate new initiatives or seek consensus on specific topics. Examples include a study to assess the European industrial view of the Japanese, and ERCIM expertise has been used in connection with the European Commission's programme for Central/East Europe countries and more recently ERCIM has been used as the facilitator for work aimed at forging closer links between the Mediterranean (particularly North Africa) countries and the Community. INRIA being a member of ERCIM was one reason why they were selected to host the European arm of the World Wide Web Consortium. The ERCIM partners have been supporting INRIA in this activity. For example, ERCIM provided help in organising the Fifth International World Wide Web Conference (Paris, May 1996) and have provided staff to INRIA to ensure a good start to their activities in Sophia Antipolis.
CLRC's participation in ERCIM
CLRC contributes to and benefits from ERCIM activities. Our contributions include the running of specialist ERCIM research groups in graphics and databases and being active in several other groups, particularly WG4 concerning the World Wide Web. We have also run workshops on technical subjects. Bob Hopgood (CLRC) chaired the ERCIM Executive Committee during the phase when the number of partners increased from 5 to 12 members. Keith Jeffery is currently the CLRC ERCIM Executive Committee member and has special responsibility for new members (the Czech Republic and Denmark now have ERCIM members as a result of Keith's activities).
The benefits include the opportunity to take part in EC-funded projects on a scale greater than might otherwise be possible, and better links with European SMEs. Many European contacts have been made through our ERCIM partners. A flow of ERCIM fellows through the CLRC helps with individual projects and in generally forging enduring links with other countries and organisations. ERCIM partnerships have produced a substantial body of refereed papers.
There is a large amount of information about ERCIM and its activities on the World Wide Web. The best starting point is the ERCIM home page at http://www-ercim.inria.fr/www-ercim.inria.fr
Keith Jeffery, DCI, CLRC
The Atlas Cray Users meeting
The Atlas Cray Users meeting held on 24 | October 1996 at the Royal Institution attracted an extremely good number of attendees, several of whom had come from the Parallel Processing Course held on the days prior to the User Meeting. A full account of the User Meeting is on the World Wide Web(WWW) (http://www.cis.rl.ac.uk/ATLAS/feb97/cray.html). Important points raised at the meeting included some concerns about grant proposals (see the article on the WWW by Prof Catlow at http://www.cis.rl.ac.uk/publications/ATLAS/feb97/catlow.html), the improved support for parallel user jobs through the 'multi' queues and forthcoming expansion of disk space. The 'Cray' user meeting now includes users of the DEC 8400 machine (Columbus) and there was a fascinating talk on earthquake modelling by Dr Shamita Das from Oxford.
Roger Evans, DCI, CLRC
TallShiP (High-level sharing for Parallel Programming) is an Engineering and Physical Sciences Research Council funded research project between The Department for Computation and Information's (DCI) Software Engineering Group and the University of Leeds, investigating the use of high-level, well-founded sharing in parallel and distributed systems. It brings Leeds' experience with scalable algorithms together with DCI's work on high-performance language design.
Many problems in distributed computing may be stated in terms of shared data - for example a large matrix being generated by several parallel processes, or a document being written cooperatively by several people. Despite this, programming environments seldom support complex shared data structures directly, often forcing programmers to use explicit low-level message passing. This can lead to very poorly-engineered applications with unduly short lifetimes.
A major problem with shared data is that applications can incur huge inefficiencies. This is partly due to attempting to maintain strict data consistency - for example giving all processes an identical view of the shared state. This level of consistency is often not required for the application to run correctly. A further source of inefficiency is the use of a single, generic representation for all shared data (or all data of a particular type). This fails to reflect the diversity of uses - often highly stylised - to which shared data is put even within a single application.
TallShiP's central "mast" is the notion of the shared abstract data type (SADT): a type which is specifically designed to work efficiently on a distributed platform. Each SADT has a number of different possible representations, having the same type signature but optimised to support different consistency models and patterns of use. This in turn allows us to easily change the representation of shared data to maximise performance as the application evolves.
Simon Dobson, DCI, CLRC
For more information please contact the author or visit http://www.cis.rl.ac.uk/proj/tallship.html
Reusing Design Information in the TORUS Project.
A typical engineering project accumulates a great deal of information about the product, and also on the progress of its development. This information could be very useful in future projects, but typically each project starts afresh, and the experience gained from previous projects is only informally used through the expertise of the personnel involved. Thus much of the knowledge gained from past projects is lost.
The CEC funded ESPRIT III project TORUS is addressing this problem by devising methods which promote the formal reuse of large sections of project information. Reuse of information is especially useful within three areas:
- Data: the technical information used in projects is often similar
- documentation: the form of documents does not vary much between projects, only the data they contain
- process: much can be gained from capturing and reusing the experience of how projects progress.
The TORUS project is using the international standard languages EXPRESS to represent data and the Standard Generalised Markup Language (SGML) to represent documents. These languages structure the information which allows the user to find and reuse large sections of relevant data and sections of documents accumulated from previous projects. However, project information needs to be presented in documentation. The Department for Computation and Information (DCI) at CLRC has been combining SGML and EXPRESS by developing a notion of prototype documents which have parameters which refer to values given in EXPRESS. In a particular project, documents are built by setting the parameters to EXPRESS values and automatically filling in the document with values taken from EXPRESS. Thus we achieve a separation of the reusable documents and the project specific information presented within those documents.
Brian Matthews, DCI, CLRC
For more information on this project see: http://www.cis.rl.ac.uk/proj/torus.html