Today the sum total of work computed on our clusters since we started in 2010 surpassed 100 years. That's the equivalent of a fairly powerful processor working flat out for a century. While we've accumulated quite a large number of cores in this time they're not continuously occupied, however this is still a 33x speedup overall, and our degree of parallelism is about two to one: on average a job runs for about 8 hours and does 19 CPU hours of work. The departments reaping the most reward from the cluster are statistical
sciences, computer science and mechanical engineering with speedups of 61x, 12x and 8x respectively.
Also of interest as per the graph below; while the rate of job submission is decreasing, the rate of hours computed increased rapidly to 400,000 hours per year in 2011 and has remained at this level ever since. This indicates that the cluster is being used more effectively and with confidence by our researchers. We hope to be able to improve on this as we move towards half a decade of providing computing resources to UCT and SAGrid.