Following the injection of the new and powerful Cray X-MP resource into the UK academic computing scene earlier this year, and then the recent transition to a Peer Review system for the use of the London and Manchester facilities, it is timely to begin to overview the Supercomputing scene with respect to the natural environment sciences. Indeed, it is only just becoming possible to identify the nature and scale of environmental science needs, and this, if nothing else, can be cited as one of the spin-offs of Peer Review: the users have had to make themselves known (to NERC, in this case).
Although other articles in the future will hopefully address the role and benefits of Supercomputing in specific scientific areas and projects, this contribution concentrates on the organisational and administrative aspects, as well as summarising the source and current level of demand for Supercomputing in the natural environment sciences. With the Natural Environment Research Council (NERC) now charged with administering the Peer Review procedures in this area, it is based on the bids submitted to two Research Grant rounds, as well as the identification of on-going projects, and the experience of some major new "community" initiatives. (The latter, incidentally, bear comparison with SERC's "Collaborative Computational Projects").
NERC is principally concerned with the earth, marine, atmospheric, terrestrial and freshwater sciences. Historically the early supercomputer applications, and now those requesting the largest resources, were concerned with modelling largely physical processes within the fields of oceanography and atmospheric science. However, requests for machine time have also been received from workers in disciplines such as hydrology, mineralogy, and geological sciences. Indeed, it can be noted that one of the biggest customers for supercomputers is the oil industry in which many applications concern natural environment science, and especially geology (as in the modelling of oil and gas reservoirs, as well as the processing of seismic exploration data).
NERC funds research in both its Institutes and in Universities (and comparable establishments). Some supercomputer use is also recorded under Studentships funded by NERC. Internally, the Institute of Oceanographic Sciences (IOS) was one of the early and significant users of the Daresbury Cray (prior to its move to ULCC). More recently, a number of major collaborative projects are being funded and administered by NERC. Two in particular, bringing together Institute and University researchers and supercomputer users, have been established as "Community" projects. As announced in "NERC News" (March 1987) these are: FRAM, the Fine Resolution Antarctic ocean Modelling project (see also article in "NERC Computing", June 1987); and the North Sea project, concerned with modelling three-dimensional transport and dispersion. In another case, a (small-c) community project requesting major supercomputer resources has been initiated as a collaboration between several University departments and workers (viz. the UK Universities Global Atmospheric Modelling Project. GAMP; see "NERC News", March 1987 for a background article on the science involved).
Unit-bids for supercomputer resources, whilst not that great in number (about forty so far) have therefore come from a broad range of sources (from individuals to communities), and have requested widely varying amounts of machine time (from 30 minutes, to 3000 hours per year). The variety of the bids has presented some procedural and administrative complications especially in this year's transitional period of rapid Cray X-MP start-up and the changeover (officially notified to NERC in April) to a Peer Review system at the London and Manchester centres. In response to this situation, the writer has been requested to fulfil a (part-time) coordination role as an intermediary between users, the peer review bodies and the super-computing centres. Not least of the problems has been the need to identify existing and potential users in order to let them know what was happening. Whilst relatively straightforward in the case of NERC Institutes, this has not been so easy in the case of the Universities where researchers and students have not normally had to refer to NERC to obtain supercomputing resources in the past. Superimposed upon this has been the evolutionary nature of new procedures and documentation being introduced by the computing centres, as well as the simultaneous establishment of new committees and management mechanisms to overview and administer the UK academic supercomputer resources.
As introduced above, the community projects (especially FRAM and GAMP) include the biggest bids for supercomputer time. Whilst, as expected in such newly constituted collaborative activities, modest resources are being used in the early months, these projects are scheduled to use up to more than 25% of the Cray X-MP resource over the next few years. In addition, FRAM and a number of other applications envisage using more than 5% of the ULCC Cray, and a variety of other bids (both approved and in the pipeline) may take requests for the UMRCC Cyber over the 10% mark.
NERC Institute users are principally from the Proudman Oceanographic Laboratory (formerly part of IOS), as well as IOS Deacon Laboratory, and the British Geological Survey. University users come from a variety of departments: Earth Sciences, Geology, Geography, Meteorology, Physics, Chemistry, Mathematics, Civil Engineering and others. Some particular centres include the Hooke Institute at Oxford (FRAM Core Team), and the Department of Meteorology at Reading (GAMP Core Team).
As we now (hopefully) enter the new steady-state regime for supercomputing, key elements in relation to the administration and organisation for NERC-science applications may be summarised as below.
Leaving aside the categories of "industrial use", and "small users" of ULCC and UMRCC (for a few hours of time allocated on an annual basis and normally bid for through the user's local computer centre), the bulk of the NERC-science applications are handled as follows. The new standard application form (ref: NS1) is used to apply for supercomputer resources at all three centres whether requiring full peer review, or falling into the lesser category of "pump-priming" (up to 0.1% of resource - a few hours - for up to three months). In the case of the former class of "large" users, the NS1 form should normally be accompanied by the relevant sections of the standard NERC Research Grant application form (ref: RG1). Proposals for research programmes which entail significant use of a supercomputer (typically over a number of years) are peer reviewed through the established NERC mechanisms. An assessment is made of both the scientific merit of the proposal and the soundness of the computational approach being adopted. University applications are peer reviewed through the Grants Committees whilst applications from NERC Institutes and large Community projects are refereed and considered by the appropriate Science Committees.
Currently there are two Research Grant rounds per annum (closing dates as for any application: 1st December and 1st July), as well as one round for Studentships (closing date: 30th November). The present round includes bids for ULCC and UMRCC resources, whilst some of the bids for the new Cray X-MP at RAL were considered in the previous round (the outcome of which was known in April/May of this year). The computer centres are appraised of bids as soon as possible after they enter the system. So far it has not been necessary to attempt the difficult task of prioritising bids across the various disciplines involved. However, as the supercomputer load builds up, and as and when guidelines are evolved for the allocation of resources between the various sciences, this may become necessary in the future.
In relation to such management and policy matters, NERC interests are represented on the Management Committees of the three supercomputing centres, whilst a representative from NERC science (Professor B Hoskins, Meteorology Department, University of Reading) is a member of the recently established Joint Policy Committee for Advanced Research Computing (JPC). NERC is additionally represented by an assessor (Mr B J Hinde, Director of NERC Scientific Services) on the Computer Board. Moreover, with a channel of reporting into both the JPC and CB, computer centre staff also meet with representatives from the Research Councils and the British Academy within a "Supercomputing Working Group".
Further information (and contacts) on these activities and any other aspects of supercomputing in the natural environment sciences can be obtained from me, Dr David Farmer, NERC Computer Services, Holbrook House, Station Road, Swindon
