Melbourne Uni steams over supercomputer critique

Victoria's auditor-general has ignited a storm over the state's $100 million life sciences computing facility after questioning whether it represented "value" for public money.

The Victorian Life Sciences Computation Initiative (VLSCI) involved the University of Melbourne, technology partner IBM and $50 million in state government funding.

It was expected to establish two x86 clusters (by IBM and SGI) and an IBM Blue Gene/P supercomputer by 2012 - the most powerful in Australia.

But an audit tabled today concluded that despite "positive early signs" from researchers, "the University cannot demonstrate that the initiative represents the most effective use of the $50 million provided by the state government".

The Auditor-General's conclusion was based on an apparent lack of scoping and cost-benefit analysis undertaken by the University.

The report was met with heavy criticism from the University's vice-chancellor Glyn Davis and from the state Department of Business and Innovation (DBI).

DBI secretary Howard Ronaldson attacked the issue of value, pointing out the department's disappointment at the "limited acknowledgement of the progress achieved" by the initiative.

"I note that you have advised the Department that the [audit] does not conclude that the VLSCI was not the most effective use of $50 million provided by the State Government," Ronaldson said.

"Rather, to paraphrase, the [audit] reflects your judgement that assurance cannot be provided from the processes observed within the scope of the audit."

Davis said the Auditor-General's finding went against advice provided to the University throughout the project.

He said the University was "aware that [the audit office] was intending to use VLSCI as a 'case study' audit... prior to the signing of the contracts with IBM".

The University had proceeded on the basis of advice from the auditor, state government and independent advisers Pitcher Partners, LEK Consulting and PricewaterhouseCoopers, he said.

Scoping the need

Acting Auditor-General Peter Frost found that the University had established a "broad need for additional life sciences computational and research capability".

However, it did not ask researchers what their "specific computational needs [were] in terms of the type of hardware required, what software they needed to run, the speed and capability of the computer equipment, or their data storage requirements".

It also provided no insight into the "amount and nature" of current life sciences research, quantify the benefits of expanded computing capacity, or define measurable outcomes.

"Without this information, the University was unable to articulate to the market the type
of products and services it required or effectively assess proposals from the market," the Auditor-General found.

"Consequently the University could not demonstrate that the acquired facility would be fit-for-purpose.

"The business case did not provide any information about the impact or the likely benefits if a smaller and cheaper computing facility was used, which effectively presented the particular PCF [peak computing facility] proposed as an 'all or nothing' proposition."

Probity concerns

The Auditor-General noted that DBI had insisted on a tender process before tipping public funds into the project.

"DBI disagreed with the proposed preferred supplier approach and required the University to use a public procurement process to find a partner," the Auditor-General stated.

"DBI noted that the University had been led by its private sector partner and a desire for a partnership with that company."

The Department's administration of the $50 million grant was determined to be "generally sound".

But the University's procurement processes were "not well planned", the audit found, which meant the institution "cannot demonstrate that its chosen procurement approach represented the most effective use of public resources".

Frost raised a number of potential probity concerns with processes but stopped short of claiming any impropriety.

"The procurement was not demonstrably impartial," he found.

"Conflicts of interest and perceptions of bias created by the University's earlier involvement with the company subsequently chosen to partner it in the VLSCI were not adequately identified and managed."

In particular, he singled out the procurement of the x86 cluster systems, alleging early involvement by "the company selected as the University's partner" led to the playing field of the bid process not being level.