ERCIM News No.45 - April 2001 [contents]
UK Grid: Management and Architecture
by John Taylor, Director General of Research Councils
The days of client/server computing are numbered. Science applications are pushing at the very limits of current technology, and in solving their problems are developing a new generation of information technologies. Increases in several orders of magnitude in processing power, bandwidth and data storage are facilitating this process.
Scientists are facing various challenges, including:
- the acquisition and curation of extremely large, extremely valuable collections of primary data.
- the need for large in-silico experiments.
- visualisation at the desktop.
- the globalisation of scientific collaborations.
Developments are moving towards an information utility, a Grid, where processing power, data storage and retrieval, data analysis software, plus all the networking to link these tools together, are behind the wall. New middleware will allow the scientist to take data and use resources from across a linked, global system without going to each resource individually or even knowing that they exist.
The emerging Grid concept sounds very attractive. But how do we go about building a Grid? How do we ensure that it meets the needs of many different user communities, without building several, incompatible Grids? How can we work within Europe and across the world to bring the new technologies into being?
In the UK, we have set up a specific e-science programme, with £ 118 Million over three years, plus extra funds to upgrade the academic network. We have deliberately divided the programme into two main sections. Firstly, there is £74 Million in specific allocations to the different Research Councils in the UK to tackle individual application science problems for which Grid solutions could provide the answer. Areas of application science include:
- Particle physics, with petabyte data flows and extremely large data storage and communication issues
- Genomics, with huge emerging databases of the human genome, and where metadata, visualisation and interoperability are crucial
- Climate change, where large in-silico modelling with many variables is particularly important
- Engineering systems design.
The second part of the UK programme is the Core e-science programme. This will be managed by one of the Research Councils on behalf of the whole scientific community, and will have £35Million of Government support, covering both scientific and industrial pump priming. The programme will be expected to pull in at least £20 Million of additional industrial support.
This core programme will invest in generic technologies scaleable networks, shared environments, access/security, metadata and so on. It will need to work iteratively with the application programmes, learning from them, providing them with new generic technology, and disseminating best practice across the disciplines. Progress in the applications will feed through to the generic, and vice-versa. Within the Core e-science programme, academic and industrial researchers will be brought together in open source joint research.
The core programme will also be responsible for expanding linkage with Grid developments internationally. It is essential that the UK programme forms part of the European and global efforts towards developing the technologies. The core programme will be actively involved in establishing international collaboration, although individual application programmes will also have specific collaborations (eg the UK particle physicists will be working with CERN).
The third strand of the UK e-science programme is a £9 Million investment in high performance computing. Added to other baseline funding, this will be used to bring sustained teraflop computing capability to the UK, and make it accessible via the Grid infrastructure.
Parallel to these developments, the UKs academic research network, JANET, has been upgraded. The backbone (Super JANET) has been raised to 2.5 Gbit/s, rising to 20 Gbit/s in 2002 and with additional increases planned. Metropolitan and Local area networks are also being upgraded, and a major boost to university infrastructure funding will allow them to increase their own internal networks.
The e-science programme is one involving many people over a number of disciplines, tackling many different and difficult problems. The overall administration of the programme is complex. The director of this Core e-science programme is an absolutely key individual. He or she will need to be an architect, a manager, a negotiator and an advocate all rolled into one. A Steering Group will bring together the interests of the application programmes and the Core e-science programme to ensure coherence across the whole UK e-science activity.
In this way, we hope to ensure that the core programme and the applications programmes have sufficient flexibility to tackle the problems, but interact in a dynamic way to ensure that we develop compatible technological solutions. It is a difficult balance, and it is ultimately up to the Chief Executives of the Research Councils and myself to ensure that it works.
Gavin Costigan - UK Department of Trade and Industry, The Office of Science and Technology
Tel: +44 20 7271 2075