National Computational Infrastructure

NCI

The Early Years (1987-1998)

ANU was the first Australian university to take the strategic decision to establish high-end computing infrastructure and services to support research. Prior to 1987, ANU researchers, like those at many other universities, undertook their research computing on general purpose mainframe systems that were shared with the administrative and enterprise computing needs of the University. During the mid-1980s, however, driven by growing internal demands for access (principally from the astronomers and chemists), and the emergence of dedicated HPC systems in peer, international universities, ANU took the decision to allocate funds from its research budget to procure its first supercomputer—a Fujitsu/Facom VP100, following an approach to the market in 1987 or 1998.  Elsewhere in Australia, at that time, the only other (known) supercomputers were operated by the national science agency, CSIRO (Control Data Cyber 205), and the Bureau of Meteorology.

From the inception of the supercomputer, ANU established a dedicated support unit, within its Information Technology Division, known as the ANU Supercomputing Facility, with a number of “academic consultants” (initially three in number), to work with researchers to implement and optimise their codes on the supercomputer system. These academic consultants, who were all research-trained, would today be referred to as “computational scientists”.

Over the years, ANUSF grew under its foundation Head, Dr Bob Gingold, and its Academic Director, initially, Prof. Don Faulkner and subsequently Prof. Dennis Evans.  The relationship with Fujitsu evolved in a number of ways through research projects undertaken jointly by Fujitsu Laboratories and ANU in areas that included computational chemistry, linear algebra, and computer vision. Of these, the first of these continues through until today as a contracted research and development activity between Fujitsu’s Technical Computing Division and ANU.

Also over the years, the Fujitsu infrastructure was progressively upgraded through a VP2200 series vector machine in 1992, and a VPP300 series vector-scalar system in 1996, with the peak performance of the installation growing from 150 MFlops through 1.25 GFlops, to 28 GFlops for the final Fujitsu system.  It is interesting to compare the performance of these very early systems with what is available in the consumer market today and to surmise what the future will hold in 20 years. In this particular instance, the second of these systems, the VP2200 has a peak performance which is slightly inferior to that of an Apple iPad2, according to benchmarking undertaken in 2011 by Prof. Jack Dongarra, of Linpack fame.

From the onset of services from the ANUSF facility, ANU made available 10% of its supercomputer system to researchers nationally. From the early 1990s, however, motivated by ANU’s example, a number of other Australia universities, either individually or through consortia, sought to establish comparable supercomputing facilities. Evident at the time were activities associated with state-based consortia in NSW and Queensland, and institutions such as the University of Melbourne, the University of Adelaide, and the University of Tasmania. In keeping with the times, there was a diversity of infrastructure installed, in many cases funded by the antecedent of the Australian Research Council, and partner organisations.  Amongst these were the Connection Machine CM2 and CM5 systems (including one at ANU within its Centre for Information Science Research), an Intel Paragon, and various SGI scalar systems, e.g., the Power Challenge series.

As may be gleaned from the description above, Australia’s approach to the development of research computing infrastructure, during the early 1990s, was quite balkanised, with the absence of a national policy to inform developments.  The practical manifestation of this was the absence of an international class supercomputer, as measured by the Top500 list. Ultimately, in 1994, the Prime Minister’s Science and Engineering Council (PMSEC) recommended a substantial investment to establish a facility of national scale. While funds of approximately $30M were set aside in 1994-95 by the Keating Government, the absence of an implementation plan saw the strategy languish for two more years.

Following the change of Government at the national government election of 1996, the financial exigencies at the time saw the pool of funds allocated to the acquisition of a national supercomputer reduced significantly. This was counterbalanced, however, by instructions from government that the Department implement a program to decide upon the implementation by the end of 1997. Accordingly, In the latter half of 1997, the Department of Employment, Education, Training and Youth Affairs (DEETYA) sought competitive bids, in effect a public tender, for the establishment of a “Centre of Excellence in High-Performance Computing and Communications”. By the close of the tendering process, five applications had been submitted: from Queensland, Western Australia, Tasmania, South Australia, and the largest of the bids which was the ONLY, ‘nearly national’ proposal led by ANU, and involving every university of NSW and South Australia, the University of Queensland, and the University of Western Australia. The absence of Victorian universities from any of the bids, however, led to the abandonment of THIS competitive approach, and the subsequent initiation of a strategic approach that led ultimately to the next phase of development, referred to as the APAC years.

In Collaboration With