30 September 2009
2009Q4 Time Allocations
The new Sun Constellation cluster, vayu, has come through the first two weeks of production use with flying colours so we will start charging for use on October 1st, coinciding with the installation of 2009Q4 grants. (Note that usage on vayu in 2009Q3 will be accounted "after the fact" for the purposes of usage reports.)
The experience so far is that vayu provides around a factor of 2 performance improvement per cpu-core over AC and XE. Some applications do better, some worse, but we will assume that 1 SU on vayu is equivalent to 2 SUs on AC/XE. To reflect this ratio the charging rate for XE will drop to 0.5 SUs per cpu-hour in normal queue with vayu providing the standard rate of 1.0 SUs per cpu-hour in normal as of October 1st. In real terms the value of an SU will be doubled by this change.
Since the number of cpus available has decreased (vayu currently has 1472 cpus compared with AC's 1920), the actual number of SUs available decreases under this change of scale. All 2009Q4 Merit Allocation Scheme grants will be scaled by 0.6616 to reflect this decrease. Note that this means the grants have actually increased by a factor of 2*0.6616 = 1.3232 in real terms. Partner share grants will be scaled only as necessary.
Any users who find they are significantly worse off due to the decommissioning of AC (ie. AC was faster than both XE and vayu for their workload) should contact email@example.com.
- 2013 International Workshop on OpenMP
- Watch the construction of Australia's largest supercomputer
- NCI supercomputer best in Australia, 24th in world
- National Computational Merit Allocation Scheme (NCMAS) committee - EOI
- Positions Vacant
- Supercomputer Boost To Australian Research
- Five New Research Highlights
- eResearch Australasia - Submissions due by 30 June 2010
No events found