Staff

Head, Academic Services

Robert Arthur Gingold, BSc Monash, PhD

Academic Consultants

Ross Howard Nobes, BSc NSW, PhD

Murray Leslie Dow, BSc Monash, PhD Tasmania (half time)

Margaret Helen Kahn, BSc Qld, PhD (half time)

David Barry Singleton, BSc Monash, PhD

Fujitsu Software Project Programmers

Roger Alan Edberg, BA Portland State, PhD

Piotr Aleksander Wielopolski, MSc Warsaw, PhD Polish Acad Sciences

James Stephen Battle, BSc Qld, PhD Qld (until 8 October)

Terry Richard John Bossomaier, BA Cantab, MSc E Anglia, PhD E Anglia (part time, 1 July-31 December)

Fujitsu Mathematics Project and Other Associated Staff

Andrew James Cleary, BSc Virginia, MAppMaths Virginia, PhD Virginia

Markus Hegland, DipMaths ETH Zurich, PhD ETH Zurich

Peter Frederick Price, BSc W Aust, PhD W Aust

David Geoffrey Green, BSc Monash, MSc Monash, PhD Dalhousie

Visualization Programmer

Drew Robert Whitehouse, BSc Qld

Programmer

Judith Helen Jenkinson, BSc, DipEd CCAE

Systems Programmer

Mathew Bock-Mun Lim, BSc

Administrator

Barbara Louise Duckworth (until 11 August)

Susan Adrienne McLean, BA (from 12 August)

Highlights of 1993 and Future Directions

Mass Storage System Acquisition

At the time of the acquisition of the VP-100 in 1987, the University recognized that the provision of a high performance computing environment for researchers required more than just the supply of computer cycles and therefore established the Academic Consultants group in the Supercomputer Facility. In the ensuing years, it became clear that it was also necessary to provide support in scientific visualization to help researchers interpret the output of the calculations. The Supercomputer Facility took the initiative at an early stage to establish the first visualization laboratory in Australia. In recent years this lead has been followed by other universities.

The remaining component required in a balanced high performance computing environment is a system for managing massive amounts of data. Although both this and visualization were initially driven by supercomputer applications, the techniques and systems required are applicable to data generated in any way. In particular, automatic remote sensors are capable of generating enormous quantities of valuable data which must be managed and analyzed.

In 1992 the Supercomputer Facility therefore initiated the acquisition of a massive data storage management system with the support of CISR and the CSC. This was approved by the University Computing Committee, in conjunction with a proposal from Mount Stromlo and Siding Spring Observatories for data storage for the MACHO project.

A considerable effort was devoted during 1993 towards making the acquisition. The system selected will be by far the most advanced system in Australia and at the leading edge internationally. In late December the base system consisting of a robotic data silo capable of holding 6000 tapes with four tape drives from StorageTek was installed and accepted. It initially has a capacity of around 1 terabyte.

In 1994 a new generation of tape technology will be installed bringing the silo's capacity to 150 to 300 terabytes and tape read/write speeds in excess of any disk system currently on campus. The data will be largely managed by StorageTek's NearNet hierarchical file migration system.

The agreement with StorageTek also foreshadows a research and development project with the ANU. This will be facilitated by the presence of the Research and Development Centre in Canberra, currently the only such laboratory outside of corporate headquarters.

The other important component in the mass storage system will be a SPARCcentre 2000 supplied by Sun Microsystems. It is planned that the system will have 10 processors, 512 Mbytes of memory and 60 Gbytes of Sun's new high-speed disk array. It is understood that this is the most powerful SPARCcentre in Australia.

The Australian Cooperative Supercomputer Facility

In 1991, the Supercomputer Facility and the Parallel Computing Research Facility jointly initiated the formation of a consortium, the Australian Cooperative Supercomputer Facility. ACSF submitted a successful proposal to the Australian Research Council for infrastructure funds under Mechanism C. In 1992, the consortium, with an increased membership, submitted a further successful proposal requesting additional funds. The total granted is $1.8 million.

The consortium members are the ANU, the Sydney Regional Centre for Parallel Computing, the Sydney Regional Scientific Visualization Facility, the South Australian Centre for Parallel Computing and Monash University. These regional bodies involve the University of Sydney, the University of New South Wales, Macquarie University, University of Technology, Sydney, the University of Adelaide, Flinders University, the Defence Science and Technology Organisation IT Division, Australian Numerical Simulations and Modelling Services (ANSTO) and the CSIRO Division of Information Technology.

The ACSF seeks to build on the support given to regional centres in recent years by the ARC and the local institutions. It highlights the need to consolidate and build upon this investment by upgrading a shared system to the highest level funds will permit. At the same time, ACSF aims at ensuring that the smaller regional systems are properly balanced and that network connections between members are of sufficient capacity. The shared system is proposed to be based on a system at the ANU. The ANU's mass storage system will also be partly shared between the partners.

Given the difficulties of funding true supercomputer systems, whether through existing ARC mechanisms or through individual university budgets, the formation of the ACSF is a major breakthrough in cooperation between the universities in high performance computing.

Connection Machine and the Time Allocation Committee

In 1993, the Parallel Computing Research Facility agreed to place fifty per cent of the time on the Connection Machines (in particular the CM-5) for `production' purposes under the allocation processes of the Time Allocation Committee (TAC) that has allocated time on the VP machines since 1987. As detailed later in this report, this began in April and has proved very successful, with demand exceeding supply. It is also interesting to note that the user base of the CM-5 is very similar to that of the VP2200 and that researchers using the CM-5 through the TAC process consumed almost all of the CM-5 resources used since April.

Information Technology Directions Statement

In 1993 the University developed an Information Technology Directions Statement to chart future directions in all areas relating to information technology. This process was conducted through a series of eight working parties and a main committee. The working party most relevant to the work of the Facility was that in Advanced High-Speed Computation Services, chaired by Professor W Levick, JCSMR. Its scope was to `review the needs of the University for advanced high-speed computation services and consider the role the University should take relative to similar organizations in Australia and worldwide'. There were some twelve members of this working party representing a variety of research areas interested in computationally demanding scientific computation. Drs Gingold and Kahn from the Facility were members of the working party.

The Working Party noted how well placed the ANU was to take a leading role in high performance computing and computational science and made a number of recommendations for future directions. These included the recommendation that the University consider funding a pool of academic posts in computational science, suggestions for enhancing organizational arrangements and changes in the career structure for Supercomputer Facility staff as well as increased educational efforts at both undergraduate and postgraduate levels.

Technological Imperatives and the Future

This year marked the end of the originally planned lifetime of the VP-100 purchased in 1987. During those years there have been marked changes in advanced computing technology, with the rise of powerful desktop workstations on one hand and the increasing acceptance by the computational science community of parallel supercomputers on the other.

While the upgrade of the VP-100 to the VP2200 in 1991 enabled the University to continue to provide a modern vector processing system to researchers for a longer period than would have otherwise been the case, it was clear at the time of the purchase of the VP2200 that by the end of 1993 the VP2200 could not be considered a true supercomputer and would be increasingly overtaken by powerful workstations. This view has been borne out by experience. While the VP2200 may still be several times faster than the most powerful workstations, when shared among many users it becomes decreasingly attractive to an individual researcher.

Many computational scientists whose work required a supercomputer in the last decade will now, if the scale of their computations has not changed, be satisfied by using modern workstations. However, it is clear that leading-edge computational science, which is modelling the world in an increasingly realistic manner, will require access to computational resources orders of magnitude higher.

Therefore, while many researchers will be content with workstation power, if there is no access to new generation supercomputers, Australian computational science will fail to tackle the leading-edge problems of the decade. We are in danger of re-entering a phase analogous to that in the early eighties, when many researchers were content with the relative power and friendliness of departmental microcomputers compared with traditional mainframes while their competitors overseas were embracing vector technology and using Cray supercomputers.

In the USA and Europe, the so-called `Grand-Challenge' problems, requiring machines with speeds between hundreds of gigaflops and teraflops, have been identified. While the introduction of terms such as `Grand Challenge' may well be more related to funding than to research imperatives, it is doubtless true that many areas of physics, chemistry, astrophysics, engineering and biosciences will require machines of this class in this decade.

The foreseeable technology demands that such powerful machines will be parallel computers. The Supercomputer Facility recognized this trend several years ago and has been actively involved in moving researchers to the parallel platforms provided by the Connection Machines and in projects on the Fujitsu AP1000 parallel computer.

Until recently, the widespread acceptance of parallel computers has been delayed by the variety of architectures and software models offered by vendors. These have had such a short life that few computational scientists have felt it worthwhile to make the considerable investment in algorithm and code development required since such efforts would be largely wasted when the machine is replaced, either by upgrade or by change of manufacturer.

However, in 1993 the situation appeared to be changing. There is now increasing consensus in the industry regarding the broad architectural features of parallel supercomputers. More importantly from the user's perspective, standards are emerging on the software environment as seen by the user. Code developed now is therefore likely to still be useful in a decade.

Nevertheless, the widespread use of parallel computers by computational scientists will require expert assistance from persons with an understanding of both the application area and parallel computing. This move will require even more assistance than the move from serial to vector machines which drove the establishment of the Academic Consultants group in the Supercomputer Facility.

The emerging architecture of parallel supercomputers opens new possibilities in their mode of use and thus could influence acquisition decisions. These machines will increasingly be collections of up to around 1000 processors, each one similar to those used in contemporary high-end workstations, but connected by a high-speed `intimate' network which adds only a small percentage to the cost of the system. This allows the machine either to be used as a centrally managed collection of workstations or for the collective power to be used as a parallel supercomputer.

With the experience gained in recent years, and some timely changes as recommended by the Advanced Computational Services Working Party, the ANU is well placed to participate in advanced computing in the last part of the century.

General

Several users of the Facility received major honours in 1993. Dr Stephen Hyde, Applied Mathematics, won the Pawsey Medal for Physics in 1993 for his work in experimental physics. Professor K Lambeck won the Charles A Whitten Medal of the American Geophysical Union and Professor John White, RSC, was elected a Fellow of the Royal Society.

Overview of the Supercomputer Facility

The Australian National University Supercomputer Facility was established following the University's purchase of a Fujitsu VP-100 Vector Processor late in 1987. The VP-100 was upgraded to a Fujitsu VP2200 in July 1991.

The Supercomputer Facility functions as a component of the University's Computing Services, but has academic links as well. The Facility is under the direction of an Academic Director (currently unfilled) and is also a constituent research group of the Centre for Information Science Research, which has responsibility for developing the University's research activities in information science and related computational disciplines.

The role of the Academic Director is to foster the use of the VP in achieving a high level of research output in existing and new areas of computationally based research, as well as to give overall academic direction to the Facility. Related day-to-day activities and management of the Facility are the responsibility of the Head, Academic Services. The Director, Computing Services carries overall responsibility for the performance of the Computer Services Centre in undertaking the `facilities management' role for the Supercomputer Facility.

Parallel supercomputers are now becoming the dominant supercomputer platforms that serve the supercomputer needs of computational scientists. There has been a convergence of technologies, with many massively parallel machines incorporating vector processing or pipelining units into each processor. At the same time, parallel machines are maturing sufficiently so that mainstream computational scientists can begin to use them. The University recognized the need to become involved with parallel processing at an early stage and through the Parallel Computing Research Facility of the Centre for Information Science Research acquired a massively parallel Connection Machine CM-2 with 16K processors from Thinking Machines Corporation and a 128 processor AP1000 from Fujitsu. In late 1992, a 32 processor Connection Machine CM-5 was installed.

The Supercomputer Facility has long recognized the need to develop expertise in parallel computing. Since 1991 the Facility has taken an active role in supporting users through involvement in courses and assisting researchers in establishing applications on the Connection Machines. The Facility has also been involved in projects on the Fujitsu AP1000.

Therefore, while the Supercomputer Facility was created with the principal objective of supporting work centred on the VP-100 vector processor, its mission now encompasses support and involvement in advanced computational science in general, including visualization, massive data storage and parallel computing.

The formal objectives of the Supercomputer Facility, which were formulated before the above-mentioned widening of the Facility's role, are:

1. to provide an easily useable and efficient supercomputer resource for research workers requiring large-scale computational facilities, particularly those whose tasks have a vector character;

2. to function as a centre of expertise and information on the use and operation of supercomputers (in particular), and on the algorithms and techniques used to apply them in various disciplines;

3. to provide collaboration with and consultancy assistance to research workers in developing applications, algorithms and codes to take fullest advantage of the supercomputer;

4. to foster the use of the supercomputer in new research areas, and to provide the expertise required to allow new users to formulate their tasks in a manner suitable for vector processing;

5. to provide a meeting ground for users and potential users, where they can, through symposia, seminars, and direct interaction, share their experiences and develop techniques that are of use in more than one application area;

6. to administer the allocation of processing time on the supercomputer;

7. to monitor the academic output of the supercomputer, and to report on and publicize its achievements;

8. to advise on developments and upgrades; and

9. to work with the Computer Services Centre to provide an efficient service.

Staff

The four members of the Academic Consultants Group play an important role with respect to most of the aims and objectives of the Supercomputer Facility listed above, in particular items 2, 3 and 4.

The primary responsibility of the Academic Consultants is in the applications area and in establishing an appropriate environment for advanced computing-based research. Contacts with researchers on campus may take a variety of forms, from short consultancies addressing particular questions from users, to extended collaborations in which a Consultant may spend several months as part of a research team implementing an application on the supercomputer. In addition, there is a Visualization Programmer, whose role is to support researchers in the increasingly important area of interpreting the output of large-scale computations.

The Facility has a Programmer who provides general programming support to the Facility and a Systems Programmer responsible for the operating system on the VP2200 and associated workstations.

The Supercomputer Facility also has major collaborative projects in place with Fujitsu Limited, Japan. One of these projects, based largely in computational chemistry and environmental modelling, employs three scientific programmers within the Facility. The Facility also manages and participates in a joint research and development project with Fujitsu in which mathematical algorithms and software for parallel-vector processors are being developed. This project employs three research fellows located in the School of Mathematical Sciences and the Computer Science Laboratory, RSPhysSE.

Management and Staff

The management arrangements for the Supercomputer Facility, which began its seventh year of operation in 1993, were unchanged. The post of Academic Director was left vacant throughout 1993, which contributed to the growing feeling that the management should be reviewed in light of developments in the University and in the technology of advanced computing. Some aspects of this were examined by the Working Party in Advanced High-Speed Computation Services, Chaired by Professor W Levick as part of the University's Information Technology Directions Study.

While operational support for the VP2200 continued to be provided by the Computer Services Centre, since 1992 the Facility has had responsibility for systems support for all its systems. The sole VP2200 systems programmer is currently assisted on a casual basis by other members of the Facility's staff who are not expert in this area. This is far from satisfactory, and so informal arrangements were put in place in 1992 so that system support staff for the PCRF and the Supercomputer Facility could act as back-up to each other. This proved ineffective in practice since the staff concerned were not located in proximity to each other. A more formal grouping and clear lines of responsibility, however, would better safeguard the large investment in advanced computing systems that the University has made.

The ANUSF Management Committee, which is no longer a formal University committee, did not meet in 1993. The members are currently:

Professor Denis Evans, Academic Director, ANUSF (Chairman)

Dr Robert Gingold, Head, Academic Services, ANUSF

Professor William Levick, FRS, John Curtin School of Medical Research

Professor Michael Barber, School of Mathematics

Dr Robin Erskine, Director, Computing Services (by invitation).

There were few staff changes in 1993. Dr James Battle, a member of the team working on the Area 3 project with Fujitsu, left in October. Dr Terry Bossamaier, CSL, RSPhysSE worked fractional-time from July for the Facility, assisting with projects in computational chemistry and in educational activities. Ms Barbara Duckworth left on maternity leave in August and was replaced by Ms Sue McLean.

Associated Staff

Dr David Green, ARC Senior Fellow in CISR and RSBS, continued to occupy an office in the Facility and to work in close association with Facility staff. His research deals with computational issues in biology. This work has included simulation studies on the Connection Machines of cellular automata, genetic algorithms and complex systems, as well as protocols for network information systems and applications to biological problems such as gene sequence alignment, evolution and landscape ecology. In 1993 he devoted much of his time to establishing the ANU Bioinformatics Facility and its high profile on the Internet.

The Facility's staff are also involved in and manage a collaborative project with Fujitsu Limited which in 1993 employed Drs Andy Cleary (SMS), Markus Hegland (SMS) and Peter Price (CSL, RSPhysSE). This project is described elsewhere in this report.

Full details of staff can be found at the front of this report.

Education, Seminars and Related Academic Activities

Eighth Biennial Computational Techniques and Applications Conference, CTAC-93

Drs Gingold and Singleton were members of the CTAC-93 organizing committee. This biennial conference was held at the ANU from 4-9 July (including two days of workshops) and attracted around 130 computationally based researchers, including many international visitors. Five staff from the Facility as well as Dr Green attended the conference.

Dr Green gave an invited paper on Nature as Computation - an Alternative Paradigm in Biocomputing and Dr Singleton gave one of the workshops on Programming Distributed Memory Multicomputers - High Performance FORTRAN in conjunction with Lindsay Hood of Thinking Machines Corporation.

Dr Singleton was also a joint editor of the refereed published proceedings of the conference.

The conference sponsored a public lecture on 6 July by Mr Trevor Robinson, Adviser to Senator Button on Information Technology entitled Supercomputers: Essential Tools or Expensive Toys?

Grand Challenge Workshop

With the assistance of Dr Terry Bossamaier, who was employed part-time during 1993 by the Facility, a half-day workshop was held on 10 September at which twelve ANU computational scientists spoke briefly on what new science could be done in their disciplines if they had access to a supercomputer capable of teraflop speeds. This was highly successful and attended by between 80 to 100 people, including a number of interstate visitors. A number of other researchers indicated that they too would have liked to have made contributions and partly as a result, Drs Bossamaier, Kahn and Singleton undertook to coordinate a document containing such contributions as well as contributions from speakers at the workshop This is expected to be completed early in 1994.

CFD Working Group

Facility members were active in an informal group of research staff and students from MSSSO, RSPhysSE and SMS which met regularly during 1993 to discuss computational fluid dynamics issues.

Supercomputer and Related Seminars

Dr Singleton gave a seminar entitled Parallel Piece-wise Parabolic Method (PPPM) for the CM-5, at the Applied Mathematics Department, University of Adelaide, 6 April.

Mr Whitehouse gave a talk to the CSIRO Division of Information Technology on 21 May entitled Visualization Production and Research at ANUSF.

The Facility also sponsored a seminar in CRES by Dr R Oglesby on Comparison of GCM Forcing to Carbon Dioxide and Solar Luminosity Changes in the CCM1 on 12 August.

Dr Green gave a number of seminars in addition to his CTAC-93 invited lecture, including:

The Chicken or the Egg - the How and Why of Gene Sequence Alignment, Robertson Symposium on Advances in Molecular Evolution, Canberra, January.

Hypermedia on the Internet, National Resource Information Centre, Canberra, March.

Cataclysms and Chaos in Natural Systems, Mathematics Department, Wollongong University, May.

Connectivity and Complexity in Landscapes and Ecosystems, Ecological Society of Australia, Canberra, October.

The Future of Biogeographic Information - Distributed Public Domain Hypermedia, Ecological Society of Australia, Canberra, October.

Mr N Snyder, a Consulting Marketing Support Representative for the IBM Engineering/Scientific National Support Center in Dallas, USA, presented a seminar on 11 March on IBM's products and plans for high performance computing.

Dr S Napier, Deputy Director of the San Diego Supercomputer Facility gave a well attended seminar on 27 October on Industrial Applications of High Performance Computing.

Dr D Cook, formerly of Rutgers University gave a seminar on 26 July concerning statistical techniques for determining relationships in three dimensional data to a group encompassing SMS, ACAT and visualization staff.

In conjunction with CISR, a series of four seminars were presented on 5-11 October by R Stevens and I Foster of the Argonne National Laboratory on High Performance Computing Systems and Architectures, The PCCM2 Scalable Parallel Climate Model, FORTRAN M and High Performance Computing Research in the USA.

VP2200 and CM-5 Courses

Dr Singleton gave a `hands-on' Connection Machine FORTRAN course for 16 people on 1-2 March and 26-27 July.

In view of the maturity of vector technology and our established user base, no VP2200 courses were offered in 1993. New users were helped as necessary on an individual basis.

The total number of persons completing the course stands at 164. A total of 45 external users have now attended the ANU for courses and workshops presented by Facility staff.

Mr Whitehouse presented demonstrations of the AVS visualization software to interested groups on campus.

Undergraduate Educational Activities

Dr Singleton gave a three day course at the University of Adelaide in April on advanced computing. The courses were attended by third year undergraduate computational mathematics students as part of a credited course. In addition to the lectures, there were several `hands-on' sessions in which the students were assisted in solving specific problems on the VP2200. The students were able to access the VP2200 through the AARNet network. This course followed a similar course given at the University of Adelaide in 1991 and 1992. The Facility would be pleased to assist with or give similar courses for ANU undergraduates.

Dr L Brewin, a former ANUSF staff member, continued to give an 18 lecture course to undergraduates at Monash University on principles and practice of advanced computing. The course used resources available through the VP2200 External Use Scheme.

Vacation Scholar

Ms Roanne Steele, from the University of Wellington, New Zealand, completed a vacation scholarship in the Facility. She worked on a project with Mr Whitehouse investigating visualization techniques and developing new software.

Outreach to Community and Industry

The Facility organized conducted tours of the supercomputer facilities for school students and the general public on 2 and 3 April as part of the Science Festival. Several hundred people visited the facilities.

With financial help from CISR, a poster was prepared on the theme of advanced computing opportunities at ANU for postgraduate students and research. As well as being distributed during the Science Festival, copies were sent to relevant departments in all Australian universities.

Dr Gingold gave a talk to the Canberra Astronomical Society on 19 August entitled Supercomputers in Astrophysics.

Ms Jenkinson gave a lecture on Advanced Computing to twenty Information Technology students from the Australian Defence Forces Academy on 24 August.

Dr Gingold was a member of the Steering Committee of the Sydney University Warren Centre project for 1993 Utility of Supercomputers in Science and Engineering. This was a major initiative involving several supercomputer sites and universities aimed at involving Australian industry and government departments in advanced computing through a number of hands-on demonstrator projects.

Publications

The following papers by Supercomputer Facility staff members and closely related staff were published or in press in 1993. Papers by other members of the ANU-Fujitsu Mathematics Project team are not listed here.

Spin-Dependent Unitary Group Approach to the Pauli-Breit Hamiltonian I: First Order Energy Level Shifts Due to Spin-Orbit Interactions, M D Gould and J S Battle, Journal of Chemical Physics, 98 (1993) 9933-9950.

Spin-Dependent Unitary Group Approach II: Derivation of Matrix Elements for Spin-Dependent Operators, M D Gould and J S Battle, Journal of Chemical Physics, 99 (1993) 5961-5975.

Spin-Dependent Unitary Group Approach to the Pauli-Breit Hamiltonian II: First Order Energy Level Shifts Due to Spin-Spin Interactions, M D Gould and J S Battle, Journal of Chemical Physics, 99 (1993) 5983-5994.

Formulae for the Spin-Dependent Unitary Group Approach, J S Battle and M D Gould, Chemical Physics Letters, 201 (1993) 284-293.

Computational Chemistry Visualization, R A Edberg, AVS'93 Conference Proceedings (1993) (peer-reviewed on video tape).

On the Consistency of Prony's Method and Related Algorithms, M H Kahn, M S Mackisack, M R Osborne and G K Smyth, Journal of Computational and Graphical Statistics, 1 (1993) 329-350.

A State Space Method for Estimating Frequencies and Dampings, M H Kahn, Journal of Statistical Computing and Simulation, 48 (1993) 101-116.

An Improved Criterion for Evaluating the Efficiency of Two-Electron Integral Algorithms,
M J Frisch, B G Johnson, P M W Gill, D J Fox and R H Nobes, Chemical Physics Letters, 206 (1993) 225-228.

Theory and Applications of Open-Shell Spin-Restricted Møller-Plesset Theory, D J Tozer,
N C Handy, R D Amos, J A Pople, R H Nobes, X Ming and H F Schaefer III, Molecular Physics, 79 (1993) 777-793.

Octaplane: A Saturated Hydrocarbon with a Remarkably Low Ionization Energy Leading to a Cation with a Planar Tetracoordinate Carbon Atom, J E Lyons, D R Rasmussen, M P McGrath, R H Nobes and L Radom, Angewandte Chemie, in press.

Gravitational Waves in General Relativity: XIV Bondi Expansions and the "Polyhomogeneity" of Scri, P T Chrusciel, M A H MacCallum and D B Singleton, Philosophical Transactions of the Royal Society of London, in press.

Proceedings of the Sixth Computational Techniques and Applications Conference (CTAC93), D E Stewart, H J Gardner and D B Singleton (1993) World Scientific, Singapore.

Some Comments on Distributed Network Databases, D G Green, in Needs and Specifications for a Biodiversity Information Network, (eds V Canhos, D Lange, B E Kirsop, S Nandi and
E Ross), United Nations Environment Programme, Nairobi (1993) 117-123.

Emergent Behaviour in Biological Systems, D G Green, in Complex Systems--From Biology to Computation, (eds D G Green and T J Bossomaier) (1993) 25-36.

Ecology and Conservation--the Role of Biological Collections, D G Green, Australian Biologist 5 (1993) 48-56.

Databasing the World, D G Green, INQUA--Commission for the Study of the Holocene, Working Group on Data-Handling Methods, 9 (1993) 12-17.

Hypermedia and Palaeoenvironmental Research, D G Green, INQUA--Commission for the Study of the Holocene, Working Group on Data-Handling Methods, 10, (1993) 11-14.

The World Wide Web, D G Green, Capital ACS Newsletter, Australia Computer Society, October-November (1993) 9-16.

Complex Systems: From Biology to Computation, D G Green and T J Bossomaier (eds) (1993) IOS Press, Amsterdam, 390pp.

FireNet--an International Network for Landscape Fire Information, D G Green, A M Gill and A C F Trevitt, Wildfire--Quarterly Bulletin of the International Association of Wildland Fire, 2 (1993) 22-30.

A Microcomputer Model for Predicting the Spread and Control of Foot and Mouth Disease in Feral Pigs, R Pech, J C McIlroy, M F Clough and D G Green, Proceedings of the 15th Vertebrate Pest Conference, Newport (1993).

Connectivity and the Evolution of Biological Systems, D G Green, Journal of Biological Systems, in press.

Databasing Diversity--a Distributed Public-Domain Approach, D G Green, Taxon, in press.

Nature as Computation--an Alternative Paradigm in Biocomputing, D G Green, Proceedings of the Eighth Computational Techniques and Applications Conference (CTAC93), (eds
D E Stewart, H J Gardner and D B Singleton) (1993) World Scientific, Singapore.

Simulation Studies of Connectivity in Ecological Systems, D G Green, Pacific Conservation Biology, in press.

The Year of the Web, D G Green, INQUA--Commission for the Study of the Holocene, Working Group on Data-Handling Methods, in press.

Getting Started with FireNet, D G Green, Wildfire--Quarterly Bulletin of the International Association of Wildland Fire, in press.

Papers Submitted for Publication

Oxirene: To Be Or Not To Be?, G Vacek, J M Galbraith, Y Yamaguchi, H F Schaefer III, R H Nobes, A P Scott and L Radom, Journal of Physical Chemistry, submitted.

On-line Resources and Information Services, D G Green, in Statistical Modelling of Quaternary Science Data: a Practical Manual. (eds T Maddy and J S Brew), submitted.

Time Series Analysis in Palaeoecology, D G Green, in Statistical Modelling of Quaternary Science Data: a Practical Manual. (eds T Maddy and J S Brew), submitted.

Internal Papers

Supercomputer Facility staff also contributed to a number of internal publications of the Program for Advanced Computations of SMS and CSL, RSPhysSE as part of the Area 4 project with Fujitsu.

Conference Attendance and Other Travel

Conferences, Workshops, Lectures and Business Trips

Dr Gingold was part of a delegation headed by the Deputy Vice-Chancellor, Professor Deane Terrell, to Japan from 9-14 May to attend a regular ANU-Fujitsu management meeting. Plans for the Area 3 project, especially the start of the new environmental modelling stream in the project, were discussed, as were plans and progress in the Area 4 parallel mathematics library project.

Dr Gingold attended a board meeting and the annual meeting of the Fujitsu International Supercomputer User Group in Santiago de Compostela, Spain on 27-28 September. He also visited ETH, Zurich on 13-15 October.

Dr Gingold visited StorageTek in Colorado on 12 November with Professor M McRobbie, CISR, to discuss research and development opportunities in the area of massive data management. He then attended Supercomputing 93 in Portland on 14-19 November.

Drs Dow, Gingold, Green, Kahn and Singleton and Ms Jenkinson attended the Eighth Biennial Computational Techniques and Applications Conference, CTAC93 held at the ANU from 4-9 July.

Dr Singleton visited Thinking Machines Corporation headquarters in November and attended a CM FORTRAN Optimizing course on 10-12 November. On this trip he also visited the Naval Research Laboratory in Washington and attended the Supercomputing 93 conference in Portland on 15-19 November.

Dr Wielopolski attended the annual Statistical Mechanics Meeting at the ANU on 11-12 November.

Dr Nobes attended a Workshop on High Performance Computing in Chemistry, National Institutes of Health, Bethesda, USA, on 6-8 March and the Third World Congress of Theoretical Organic Chemists, Toyohashi, Japan, on 18-24 July. This was followed by a meeting in Tokyo with Fujitsu Ltd to discuss technical aspects of the Area 3 project.

Mr Lim attended the Australian Unix User Group, AUUG'93 conference in Sydney from 29-30 September and also gave a talk on UNIX Resource Management Systems to the Fourth Annual Canberra meeting of the Australian UNIX Users Group on 17 February in which he described the RASH system developed for the VP2200.

Dr Green was a convenor for the Robertson Symposium on Advances in Molecular Evolution held in Canberra in January and also a convenor for the Australian Academy of Science Boden Conference: Molecular Systematics of Gene Families which was held at Thredbo in February. He also attended the Ecological Society of Australia meeting in Canberra in October and visited Hobart in August to install World Wide Web for the Tasmanian Parks and Wildlife Service.

Mr Whitehouse visited the Sydney Regional Scientific Visualization Laboratory in October.

Collaboration with Industry

Area 3 Collaborative Software Project with Fujitsu Japan

The Area 3 collaborative software development project with Fujitsu Japan was established in 1989 with the aim of porting and optimizing computational chemistry software on the VP-100. After a highly successful first year, the scope of the project was widened to include mathematical software and, in 1992, graphics software.

In 1993, a new sub-project in environmental modelling was begun with Fujitsu's Research Centre for Computational Science. The CCM1 climate model from the National Center for Atmospheric Research, USA was established on the AP1000 parallel computer provided to the University by Fujitsu Laboratories as part of a joint project with the Department of Computer Science and the Computer Science Laboratory. This work was done in conjunction with Gavin Michael and other members of the Department of Computer Science. By the end of the year, it was clear that establishing this program on the AP1000 had strenuously but successfully tested many components of the AP1000 environment being built at ANU.

As well as some porting and optimization activities, work was continued on implementing computational chemistry algorithms and software on the massively parallel AP1000. Planning and actual porting was begun for establishing computational chemistry packages on the VPP500 supercomputer. Other aspects of the project in 1993 included investigations of the comparative functionality and performance of the current computational chemistry software base and visualizations of results in current areas of research. A video tape was almost complete at year end describing computational chemistry and the Area 3 project at an executive level.

In this the fifth year of the project, three full-time staff were employed (Dr James Battle, Dr Roger Edberg, and Dr Piotr Wielopolski). The scientific and programming management of the project was under the control of Dr Ross Nobes.

A large number of computational chemistry packages, all of academic interest, and many also of commercial interest, have been completed in this project and form the basis of a comprehensive list of such software available on Fujitsu machines. Several of these packages are in production use at ANU and are also heavily used by external users.

Under the agreement for the project, funds were also provided in return for machine time used during the development work; these funds have been used to upgrade the hardware and system software on the Fujitsu systems, return funds to the University Computing Committee and to recompense the Computer Services Centre for facilities management.

During the year there were many contacts between Fujitsu Japan and ANU relating to this project, including visits by Dr R Gingold and Dr Nobes to Japan and visits to the ANU by Fujitsu staff.

Area 4 Parallel-Vector Mathematics Subroutine Library Project with Fujitsu Japan

This project completed its first year and the University and Fujitsu agreed to extend it for a further year at an expanded level. The project is now squarely focussed on researching and developing mathematical library algorithms and code for Fujitsu's new VPP500 supercomputer, effectively the fastest machine in the world currently.

The Supercomputer Facility is managing this project and four of the staff are actively involved in the work. Academic direction of the project is under Professor R Brent, CSL, RSPhysSE and Professor M Osborne, Program in Advanced Computation, CMA, SMS.

Three research fellow positions are now funded under this project. Dr Andrew Cleary, formerly of Sandia National Laboratories, USA, arrived at the ANU late in 1991 to begin work on the project. Dr Markus Hegland, formerly of ETH, Switzerland joined the project in January 1993. Dr Peter Price, who previously had a half-time appointment with ANUSF, worked half-time on the project complementing his other half-time position working in developing numerical algorithms in the Area 2 AP1000 project. A full-time post was advertised during the year, but the successful applicant eventually declined to take up the appointment. It was expected that an alternative would be appointed at the start of 1994. Drs Cleary and Hegland are located in the Program in Advanced Computation, CMA, SMS and Dr Price is located in the Computer Sciences Laboratory, RSPhysSE.

The project has excited considerable interest among staff of the SMS, and a number of others, including Dr S Roberts, Dr D Stewart and Dr Z Leyk have been involved. Dr M Dow, Ms J Jenkinson, Dr M Kahn and Dr D Singleton of the ANUSF have also been actively involved in the project.

One Australia Yacht Design

During the year, the University entered into an agreement with Sun Microsystems to supply supercomputer resources and support to assist the design by Fluid Thinking Pty Ltd of a challenger for the America's Cup by the One Australia Consortium led by John Bertram. Work began late in the year on this project which is thought to be the largest industrial supercomputer project undertaken in Australia.

ANU-Ford Collaboration

The Facility is a partner in a collaborative project that is being established with Ford in Melbourne by the Department of Engineering, The Faculties. The collaboration is aimed at improving the quality and manufacturing of automotive components while reducing the cost and weight.

Visitors to the Supercomputer Facility

During 1993 there was a visit by staff of Fujitsu Japan for a regular management meeting for the joint projects between ANU and Fujitsu. Visitors included:

Dr Y Kosaka, Engineer in Chief, Systems Business Promotion Group 1-2 February

Mr N Tahara, Research Centre for Computational Science 1-2 February

Mr T Okuzumi, Oceania and Canada Division 1-2 February

As well as a number of visits from external users of the VP and from industrial representatives, the following persons visited the Facility in 1993:

Prof M Arbib, University of Southern California 8 February

Mr H Fukuda, Deputy Director, Industrial Electronics Division, MITI, Japan 15 February

Dr R Garner, IBM Cornell 22 February

Dr A Beswick, Department of Primary Industry, QLD 25 February

Mr B Wood, Minister for Education and Training, ACT Government 25 May

Mr L Bray, Ernst and Young
31 May

Dr R Oglesby, Purdue University 23 July - 20 August

Mr A MacDonald, Director of Information Services, University of Calgary 28 July

In addition, a delegation led by Mr E Izumida from the Fujitsu Center Association, Japan visited on 8 March.

Hardware, Software and System Performance

The VP Hardware, Software and Operating Environment

The Fujitsu VP2200 is a water-cooled 64-bit vector processor with a peak speed of 1.25 gigaflops. It has a clock cycle time of 3.2 ns (6.4 ns scalar cycle time), 32 kbytes of vector registers and four words-per-cycle bandwidth to memory. The main memory is 512 Mbytes and there is 1 Gbyte of SSU (secondary memory). During 1993 it was configured with around 35 Gbytes of disk space, including 15 Gbytes on loan from Fujitsu Australia Limited.

The VP2200 runs Fujitsu's System V Release 4 UNIX (UXP/M) and is available both for batch and interactive use. Network access is through TCP/IP connections onto the campus Ethernet, AARNet and the Internet. An FDDI connection is planned for early 1994.

Both a vectorizing FORTRAN compiler and a vectorizing C compiler are installed together with third-party software packages.

The VP2200 system and its software are now quite mature and few changes occurred in 1993. The system was available to users for 98.7 per cent of the year. The monthly availability is illustrated in the following figure.

Click here for Picture Although usage of packages is not high at ANU since many researchers are developing their own codes, the packaged software base is gradually growing. In 1992, the University Computing Committee allocated funds to the Facility to purchase major software packages if similar funds were provided by potential users of the packages. The DISCOVER biomolecular modelling package from Biosym was purchased in 1992 from this fund in conjunction with the Centre for Molecular Structure and Function. In 1993, the balance of the funds were used to assist with acquiring a site-wide license for AVS, the leading visualization package. The Facility was unsuccessful in attracting funds for 1994 through this mechanism.

Packages and libraries on the VP2200 include:

Chemistry Mathematics Graphics Biological Sciences

CADPAC IMSL 1.1 PGPLOT DISCOVER

ACES II NAG 14 NCAR AMBER

GAMESS ELLPACK HDF X-PLOR

MOPAC 93 LAPACK

AMPAC 2.1 BLAS Environmental Sciences

CRYSTAL 92 ITPACKV CCM1

GAUSSIAN 92 SSLII

MM2/MM3/MINP

SPARTAN

COLUMBUS

MOLPRO 92

WIEN

The RASH System on the VP2200

The UNIX operating system and its associated accounting mechanisms does not satisfy the Time Allocation Committee's requirement to make grants on a project basis and at the same time allow users, who may wish to work on more than one project, to have a single individual identity on the VP2200. The Supercomputer Facility therefore has designed and implemented a shell, the Resource Accounting Shell (RASH), to provide these features. RASH has attracted considerable interest at international VP sites and, following a suggestion from Fujitsu Japan, the Supercomputer Facility now markets RASH to Fujitsu and Siemens sites. The High Performance Computing Center in Calgary, Canada is the first RASH customer.

Other ANUSF systems

The Facility also makes generally available several graphics workstations in the Visualization Laboratory. These and the software environment are described in the relevant section of this report. In addition, a HP9000/735 workstation with a peak speed of 200 Mflops is available on a limited basis. The MATLAB package is installed on this machine.

Mass Data Storage System

In late December a robotic cartridge data silo supplied by StorageTek was installed and accepted. The ACS4400 initially has approximately 1 terabyte capacity and is equipped with four 18 track tape drives. These are expected to be upgraded to helical scan `Redwood' drives with speeds in excess of 10 Mbytes per second and tape capacities such that the total amount of possible storage will be in the 150 to 300 terabyte range. The silo is initially attached to a Sun 690 that formerly was used as a front-end to the decommissioned CM-2, but this is to be replaced by an eight to ten processor SPARCcentre 2000 with 60 Gbytes of disk array early in 1994. The system is expected to progressively go into production use in 1994.

Connection Machines

In 1993, the Facility increasingly spent more time supporting the Connection Machines of the Parallel Computing Research Facility. As explained elsewhere in this report, a major part of CM-5 resources were subject to allocation by the Time Allocation Committee from April.

The 16,384 processor CM-2 was decommissioned in November. The CM-5, installed in June 1992 with 32 nodes, was upgraded in November 1992 with vector units and now has a peak speed of 4 Gflops and 1 Gbytes of memory. It has a front-end Sun 4/690 with four processors, 256 Mbytes of memory and 8 Gbytes of disk. The CM-5 has a FORTRAN90-like compiler, C* (an extension of the C language), *LISP, the CMSSL scientific subroutine library, the CMX11 graphics library and miscellaneous support tools.

During the period covered by the Time Allocation Committee the CM-5 was operational approximately 97 per cent of the time.

Scientific Visualization Laboratory

Scientific visualization is the generic name given to techniques that use images and animations to interpret scientific data. Although not restricted to supercomputing applications, the field has largely been driven by the difficulty of interpreting the enormous amounts of data produced by many supercomputer applications. Such data are often not only massive but multi-dimensional and time dependent and so conventional graphics techniques are often completely inadequate.

The Facility's initiative in scientific visualization began in 1990 with the appointment of a Visualization Programmer, initially on a two-year basis, but now extended indefinitely. The main aim of the initiative is to provide software support and general visualization expertise and advice to users of advanced computers. The major visualization tool, a Silicon Graphics VGX workstation, is heavily used by over 30 registered users. In many cases researchers have used the Visualization Laboratory's facilities to help decide upon the optimum solution to their visualization requirements before purchasing personal and departmental systems.

The Visualization Laboratory in the Leonard Huxley Building is accessible at all hours to approved users through the CARDAX security system. Currently there are 65 users who have access. The Visualization Laboratory houses the following equipment:

* a Silicon Graphics 4D/210 workstation with the VGX graphics option
* a DEC 5000/240 workstation with PXG turbo graphics
* the frame buffer of the CM-2
* the Sun graphics workstations which are the front-end consoles of the CM-2
* a Sony laser-disk system for storing and playing back images
* a Chromatek scan converter

(Equipment associated with the CM-2 was removed when the CM-2 was decommissioned.)

In addition, a large amount of video equipment is generously on loan jointly to the Visualization Laboratory and CSIRO Division of Information Technology by the National Science and Technology Centre. This equipment is housed in the Laboratory and together with the above and other equipment has enabled the production of high-quality video tapes displaying data obtained through supercomputer simulations or from experiments.

During 1993 it became clear that the demand for visualization services are greater than can be met by a single Visualization Programmer whose time has become increasingly fragmented between investigating new products, helping users, making video tapes and education. As well as numerous consultations and demonstrations, seven videotapes were made for users during the year. Software for high performance volume rendering was developed for the CM-5 and AP1000 and internationally available modules written for the Silicon Graphics Explorer system.

During 1993, the Laboratory organized a campus-wide license for the AVS visualization system which was installed on the Silicon Graphics, DEC and HP735 workstations, the CM-5 front-end and machines in the Department of Computer Science, MSSSO and RSPhysSE. This software has been used extensively for projects in chemistry, earth sciences and astrophysics.

The MSI ChemistryViewer software was installed on the DEC AVS system and used extensively by staff from the Research School of Chemistry and as part of the collaborative project with Fujitsu.

Other software installed in the visualization laboratory includes Explorer 2.2 from Silicon Graphics (the Facility was a beta test site for this) and Scian.

Late in the year, much work was done by Dr Kahn and Mr Whitehouse on preparing a new video tape illustrating the work of several ANU researchers and highlighting advanced computing at ANU. This should be complete early in 1994. Copies will be available upon request to the ANUSF.

Dr Edberg had a video segment on computational chemistry accepted for `publication' at the AVS'93 Annual Conference in Orlando, Florida in May.

External Use Scheme for Australian Universities

Following the acquisition of the VP-100 in 1987, the University decided to provide ten per cent of the capacity of the vector supercomputers free of charge to researchers at other Australian universities. This amounts to around 750 hours per year, which at commercial rates (which some universities have paid in the past to bureaux for supercomputer time) would cost approximately $1 million per year.

Until this year, the ANU did not involve itself in deciding the relative merits of projects from external universities. Over the first five years of this External Use Scheme, a formula provided by the Australian Vice-Chancellors' Committee was used to allocate time in bulk to each participating university which then informed the ANU of its own sub-allocations to individual projects.

By the end of 1992, it was clear that the system needed modification. The number of projects grew to be too large, with the result that each project received an inadequate grant. For example, in 1992, there were 81 projects registered with over 109 users at 24 universities (not counting a further 35 course-work accounts). This is an average of only 10 hours per project. This is grossly inadequate for serious production work and was perceived as such by users and potential users.

Therefore, after discussions with a number of external users and coordinators, it was decided to alter the scheme from April 1993 with the aim of encouraging a small number of larger-scale projects. Smaller projects that made use of special features on the VP such as software packages, large memory and rapid turnaround of urgent work will continue to be encouraged, and resources for educational activities are still provided.

It was decided to subject most allocations of the resources reserved for the External Use Scheme to the University's Time Allocation Committee. In view of the small overall resources being allocated, it was decided that it was not worthwhile expanding the Committee to include external representatives at this stage. Vice-Chancellors, external coordinators and users were notified of these planned changes well in advance of the introduction of these new arrangements. Some modifications were made in light of responses.

At the March meeting of the Time Allocation Committee, 18 applications were received, requesting three times the resources available. Guided by international and national referees, the Committee determined that almost all of the proposals were of an exceptionally high standard and found it very difficult to reduce the number of supported projects. Fourteen were given the maximum support possible (after weighting in light of referee's comments) and the remainder were given small provisional grants. The Committee decided to over-allocate the resource available by 50 per cent so as to encourage users to take up their grants. At the September meeting, twenty proposals were received and similar policies were followed in making the allocations.

Fifteen of the external projects were supported independently by the Australian Research Council.

Table 1 lists the external projects which were supported by the Time Allocation Committee. Seventeen projects involving 35 researchers are listed.

During the nine months represented in the tables, total usage was 557 Service Units which is in line with the resources available. Omitted from the table are several projects which were given small provisional grants and which, not surprisingly, made little or no use of the grant. Detailed descriptions of the projects can be found in Appendix B.

In the three months of 1993 before the new scheme was introduced, external users consumed 233 Service Units, bringing the total to the year to 790 which is slightly in excess of the amount formally made available under the scheme.

Despite the success of the new scheme, it is still clear that the amount of resources being offered to each project is grossly inadequate in relation to the excellent research work proposed. Some users did not take up their grant as a result. It is feared that unless a further reduction in the number of supported projects is made, grants will again be too small.

Using resources under this scheme, Dr L Brewin, Mathematics Department, Monash University, gave an 18 lecture course to honours students at Monash University on the principles and practice of advanced computing. The course was assessed by worked examples. Courses taught by ANUSF staff at the University of Adelaide (which are described elsewhere in this report) also used accounts under the External Use Scheme.

Table 1 -External University Principal Investigators - Grants and Usage on the VP2200

Time Allocation Committee and ANU Usage

The Time Allocation Committee (TAC) has had the responsibility for allocating time on the Fujitsu vector processors since 1987. In March 1993 it was agreed between ANUSF and PCRF that the TAC would take responsibility for making grants of production time on the Connection Machines. In the event, only the CM-5 has been subject to the TAC procedures. Formally, the PCRF agreed to 50 per cent of the CM-5 being allocated in this way, though in practice the actual usage by TAC approved projects was far greater.

The Committee's policy is that all but small start-up allocations of time to local researchers should be subject to a peer-review process. This is in contrast to other centralized computer resources at the ANU where there are no such procedures. Applicants are required to provide an academic justification for the amount of time requested, as well as information by which the suitability of the computer system for the proposed research can be judged. The Chairman of the Committee obtains reports on proposals from referees throughout Australia and in many cases from overseas. New projects and continuing projects that make large requests for resources are routinely refereed. The Committee makes grants of time on the basis of these reports and on the effective use of previous grants. Since the time between meetings of the Committee is quite long, procedures are implemented to assist researchers who wish to commence projects that evolve between meetings of the Committee.

Not only do ANU researchers make full use of the available time on the systems, but the rigorous refereeing and review by the Time Allocation Committee ensures that this work is of the highest standard. Furthermore, as demonstrated in last year's Annual Report, exceptionally efficient use is made of the VP hardware. It is harder to determine the efficiency of use of the CM-5, but preliminary examination indicates that at least the largest projects are making efficient use of the system.

The Committee met twice during the year (on 23 March to allocate time from April to September and on 20 September to allocate time from October to March 1994).

Membership of the Time Allocation Committee

Members of the Committee are academics appointed on a two-year rotating basis by the Chairman. The Chairman is appointed by the Vice-Chancellor. During 1993, members were:

Professor W R Levick, FRS, John Curtin School of Medical Research (Chairman)

Professor M A Bennett, Research School of Chemistry

Dr G V Bicknell, Mount Stromlo and Siding Spring Observatories

Prof R P Brent, CSL, Research School of Physical Sciences and Engineering

Dr R A Gingold, ANUSF (non-voting member)

Internal VP2200 Allocations

Since the VP2200 user base is quite mature, a new policy was implemented in 1993 under which users of large amounts of VP2200 time were not able to request substantially more than their previous grant. This made it easier to manage the allocation process and to introduce several new large users.

At each meeting the TAC received about 12 proposals for new projects on the VP2200. During the year, there have been 161 researchers associated with ANU projects on the VP2200, comprising 57 Principal Investigators and 94 co-investigators, of whom some 123 have individual user accounts on the system. (The totals obtained by summing the columns headed `Number of Researchers' in the following tables are greater than 161 because some researchers work on more than one project.)

A total of 91 internal VP2200 projects were approved by the Time Allocation Committee in 1993. Grants to ANU projects amounted to 9148 Service Units, and the total CPU time used by ANU projects was 6561 hours. (A Service Unit is related to cpu use and the priority of the job; at normal priorities, one Service Unit corresponds to one hour of cpu time.)

Some 679 hours of `bonus time' were used in 1993. Bonus jobs run at low priority and should only have access to the CPU when other jobs are performing data transfers to disk, etc. Projects can access `bonus time' when they have exhausted their allocations.

The remaining time was used by external users, for systems development and for the Fujitsu collaborative software development projects and similar activities.

Table 2 presents a summary of the number of researchers and time allocations by research field. Table 3 lists grants and usage by project. A detailed description of each project is presented in Appendix A.

External VP2200 Allocations

As outlined in the External Use Scheme section of this annual report, the TAC took responsibility for allocating the bulk of external usage of the VP2200 as from April 1993. Details of the outcome and usage are presented in that section.

Internal CM-5 Allocations

The TAC received 23 applications for projects on the Connection Machines at its March meeting. The total resources requested were 50 per cent over that formally available. With the permission of the PCRF, 75 per cent of the machine was allocated under the assumption that on a new system some users will not make use of their grants. As the figures in Table 4 show, usage was nevertheless high. At the September meeting, demand was at the same level and 25 projects were supported. A total of 32 projects and 55 users were active on the CM-5 using TAC grants. There were 11 projects, involving 74 researchers, which were granted time on both the VP2200 and CM-5 (such projects are marked by an asterisk in the tables). The Service Units used totalled 2711, with an additional 1042 hours of bonus time being used.

The user base of the CM-5 as allocated by the TAC is quite similar to that of the VP2200. Detailed descriptions of each project are presented in Appendix A.

Use of the CM-5 by non-TAC users was quite small, amounting to only 429 hours in the period covered by the TAC. Some 106 hours of this was used by a project awaiting approval by the TAC. There is no evidence that access for non-TAC users was disadvantaged by the TAC approved projects.

Although the total time used does not add up to nine months of wall-clock time, usage of the CM-5 has been quite heavy. The system suffered a number of failures during the year (particularly in October) and some time is lost each day as the system changes over from night-time mode to day-time mode. As a result the system was available to users for around only 93 per cent of the time. The accompanying figure indicates the usage as a percentage of the approximate time available in each month since April.

Figure 2 - CM-5 Usage by TAC Projects as Percentage of Machine Availablity"

Table 2 - Distribution of ANU VP2200 Time Allocations by Research Field, 1993

Table 3 - ANU Principal Investigators - Grants and Usage on the VP2200

Table 4 - ANU Principal Investigators - Grants and Usage on the CM5