THE AUSTRALIAN NATIONAL UNIVERSITY
POLICY OF THE SUPERCOMPUTER TIME ALLOCATION COMMITTEE

The criteria for resource allocation used by the Supercomputer Time Allocation Committee are:

1. Academic quality, as assessed by referees using as guidelines:
   • the originality of the proposed work;

   • the potential of the work to generate new knowledge in an important area;

   • the comparative scientific merits of the work within its discipline;

   • the experience and demonstrated research capacity of the applicant;

   • the reasonableness of the requested resource allocation.

2. The track record of efficient usage and achievement in the case of continuing grants.

Large requests

Local, national and international referees may be contacted, particularly for proposals making large requests. Investigators may nominate referees. These referees may or may not be contacted at the Chairman's discretion. Investigators may, however, nominate referees that the Chairman should not contact. Continuing projects may be refereed at intervals at the discretion of the Chairman.

Policy for new large projects:

When considering a new large project in the absence of a relevant track-record, the Committee generally makes a provisional smaller grant and releases the balance of the grant when the investigator can show evidence of reasonable progress.

The committee, whose Chairman is nominated by the Vice-Chancellor, is currently composed as follows:

Professor Denis Evans (Chairman)
Professor M A Bennett (RSC)
Dr G Davies (RSES)
Dr S Marcelja (RSPhySE)
Dr B P Molinari (FEIT)

LARGE USERS - PLEASE NOTE:

Policy for 'large users'

For some time the STAC has been concerned that the size of the requests by users of large amounts of time has become unrealistic compared with the resources available and the grants actually made. These unrealistic requests have two deleterious effects. First, referees are presented with apparently excessive requests which does not reflect well on the individuals concerned or on the University. Second, it is unfair to users who submit realistic requests.

A new policy was introduced successfully in 1993 to ameliorate this problem. In essence, unless extraordinary reasons are given, the committee will only entertain modest increases beyond the previous allocation actually granted for continuing projects that are classified as 'large'. (See below, for the definition of 'large'.)

While this apparently introduces a conservative bias into the allocation process, given the mature nature of the user base, the STAC does not believe that this is an overwhelming objection. New projects, including 'large' ones, and the growth of existing projects into 'large' projects will continue to be encouraged. New 'large projects' will continue to be handled as per the existing policy outlined overleaf.

The existing policy for handling demand in excess of supply will continue to be followed and hence grants to new 'large projects' will largely be at the expense of existing 'large users'. If the general level of demand falls, or the application of Criteria outlined above becomes more favourable, the grants allocated to continuing large projects will be increased.

Implementation:

When grants are invited, we will:

• notify Principal Investigators if their existing projects are classified as 'large'.

• inform each PI of their previous grant for each of their projects.

• for any single project, the request is for no more than the previous grant.

• the aggregate request over all their projects classified as belonging to a single 'large' project group, should be no greater than their previous total allocation.

PROCEDURE USED WHEN TOTAL REQUESTS EXCEED RESOURCES AVAILABLE:

These details are provided for those interested and generally affect only 'large users'.

The total requests for resources, even after the above procedures have been followed, have been up to twice the time available. This excess request arises largely from a small number of large projects. The STAC has developed the following mechanism to handle this:

1. For all projects except 'large projects' (see below), grants are made by applying discount factors based on the criteria described above to the request for service units made by the applicant.

2. Projects, or groups of projects with related investigators, which request a large number of service units are identified as 'large project-suites'. The number of service units corresponding to 'large' is at the discretion of the Committee and it is considered wise not to publish this figure. The procedure for allocating resources to 'large project' ensures that there is no discontinuity in allocations between proposals that are deemed to be 'small' or 'large'.

3. For new projects falling within the category of a 'large project-suite', the number of service units requested by the applicant will be discounted on the basis of criteria 1, 2 and 3 above and combined with the continuing 'large project-suites' for establishing the final allocation of resources.

4. For continuing projects that are part of 'large project-suites', applicants are required to follow the new policy detailed below when specifying the number of service units requested for such projects.

5. 'Large project-suites' share in the balance of resources left after satisfying all other applications. Each such suite is first granted, free of an additional 'large user' discount, the maximum 'small project' allocation . This 'large user' discount is applied to the balance of the suite's total notional grant after step C and is proportional to the remaining resources available divided by the remaining total of notional grants summed over all 'large project-suites'. Finally, grants are made to individual projects in each suite.