National Computational Infrastructure

NCI

The APAC Years (1999–2006)

Early in 1998, following the abandonment of the competitive bidding process, on the advice of Departmental officials, the Minister for DEETYA, Dr David Kemp, put in train a process to develop a more inclusive, large-scale organisational model to create a genuinely national infrastructure. Accordingly, from early 1998, an advisory board, referred to as the Interim Board of the National High-Performance Computing Centre (chaired by Dr Michael Sargent), with representation from every state, and each of the research funding agencies was put in place, and tasked with the development of a proposal for a suitably national model to put to the Minister for funding. What emerged from the deliberations of the Interim Board, advised by an expert national committee, together with independent advice from the Director of the then (US) National Partnership for Advanced Computational Infrastructure, Dr Sid Karin, were recommendations for a framework for the Australian Partnership for Advanced Computing (APAC), the objectives of which were to:

  • Fund a peak facility at ANU, underpinned by a partnership model to facilitate high-performance computing activity in every state, along with
  • Expertise and education programs that could operate, both centrally and nationally.

These recommendations were accepted by the Minister, Senator Amanda Vanstone, who allocated $19.5M to initiate the APAC program.  What was less well known is that the bulk of the initial funding for APAC was obtained from the Australian Research Council, which at the time was operated as a program with the Department. The decision to fund an out-of-program activity, such as APAC, from the ARC led to questions about the independence of the ARC, and contributed ultimately to the decision to establish the ARC as an independent authority with its own budget.

The recommendation to site the APAC National Facility at ANU was based on the University’s leadership of, and its considerable previous investments in, HPC infrastructure, together with its willingness to share access to its facilities. The decision to host APAC from ANU, however, was rooted in the strong support by ANU’s then Vice-Chancellor, Prof. Deane Terrell, that ANU could, and should, play a leading role in such national initiatives. ANU’s successful carriage of the governance and management of APAC owes much to Professor Terrell, and to Professor Robin Stanton (then Pro Vice-Chancellor of ANU), who anchored the responsibilities of the host organisation.

Early in 1999, the Board of APAC was established under the chairmanship of Professor David Beanland, and the appointment of the foundation (and sole) Executive Director, Professor John O’Callaghan followed soon thereafter. APAC was formally launched late in 1999, having initiated the process by which the:

  • Partnership could be established, through proposals from organisations, and consortia of organisations;
  • National Facility, i.e., the national peak system, could be established at ANU.

By 2000-01, APAC as a “partnership of partnerships” was largely in place, with the membership being:

  • Australian National University, as Host;
  • Commonwealth Scientific and Industrial Research Organisation;
  • Australian Centre for Advanced Computing and Communications (ac3), a consortium representing the universities of NSW;
  • Queensland Parallel Supercomputing Foundation (QPSF, later QCIF), a consortium representing the universities of Queensland;
  • South Australian Partnership for Advanced Computing, a consortium representing the universities of South Australia;
  • Tasmanian Partnership for Advanced Computing (TPAC), a consortium led by the University of Tasmania;
  • Victorian Partnership for Advanced Computing (VPAC), a consortium of Victorian universities;
  • The Pawsey Supercomputing Centre, a consortium of CSIRO and the universities of Western Australia.

The National Facility was ultimately commissioned early in 2001, following some early delays in the procurement process. With the installation of a Compaq/HP Alphaserver SC system of 1 TFlop capacity, Australia, at last, had an internationally competitive system, benchmarked at rank 31 in its debut on the Top500 list.

With the establishment of services, APAC operated in two phases, respectively from 2001–04, and from 2004–07.  Its first phase, funded largely by monies from the ARC, comprised the:

  • Establishment of services from the National Facility—which were delivered from the ANU Supercomputing Facility;
  • Building of national expertise in, and the uptake of, the use of advanced computing in research, through the partnership.

Following the commencement of services from the National Facility, ANU decommissioned its ageing Fujitsu system, and subsequently serviced all of its high-end computing requirements from the National Facility, through a substantial access share which it paid for through a combination of cash and in-kind contributions, embodied in the hosting arrangements for the NF system.  Each of the APAC partners, subscribed to an access share of the NF system (although, none as large as ANU), complementing their local facilities. The largest share of the NF (approximately 50%) was allocated to the APAC Merit Allocation Scheme, an open access allocation process, rooted in ANU’s earlier processes, and strongly aligned with research merit and computational suitability.

APAC also established capacity and capability programs in each of its partners, through its Centres of Expertise programs (in which each partner took leadership in a dedicated field or application area), as well as through its Computational Tools and Techniques program.

The second phase of APAC operations (2004–07) were funded through the System Infrastructure Initiatives of the Commonwealth Government.  This entailed the:

  • Refreshing of the National Facility peak system; and
  • Nurturing of a nascent national eResearch program nationally through its Grid and Collaboration Tools program, and the beginnings of a data and storage program.

With regard to the former, the Compaq/HP system was replaced in 2005 by a 14 TFlop SGI Altix 3700 system, which ranked as 26 on its debut on the Top500 list.

Throughout the APAC years, operational services for the peak facility were provided through the ANU Supercomputing Facility, which was accountable to Executive Director of APAC through the APAC Agreement. During the early years, ANUSF was led by its foundation Head, Dr Bob Gingold, and subsequently by his successor, Dr Ben Evans, following Dr Gingold’s retirement.

Towards the end of this second phase of APAC, a Department of Education, Science and Tourism (DEST) commissioned a review of APAC in advance of further infrastructure investments that were to be conceptualised under a new, collaborative framework, known eventually as NCRIS (National Collaborative Research Infrastructure Strategy). An international review noted the considerable success of the APAC program, including the high-quality of service from, and the internationally competitive ranking of, the peak system, the cultivation of a national community of researchers and support specialists, the embedding of high-level computing expertise across a bread of application areas, the nurturing of grid expertise and collaboration tools, and the fostering of engagement and commitment from government at state and federal level.

With the implementation of NCRIS from 2007 onwards, the Australian Government set in place a funding allocation to promote the uptake of eResearch techniques across the span of research, in a manner broader to that of the eScience programs in some other nations.  While the APAC model had been a considerable success, the strategic framework for Platforms for Collaboration was conceptualised differently, with distinct, investment programs for:

  • Interoperability and Collaboration Infrastructure—through the Australian Research Collaboration Service (ARCS),
  • Data Infrastructure — through the Australian National Data Service (ANDS),
  • High-performance Computing — through the National Computational Infrastructure (NCI),

in addition to other, smaller investments in authentication and authorisation, and networking.  These are referred to in the following section, which covers the NCI years.

In Collaboration With