My photo

Bill St. Arnaud is a consultant and research engineer who works with clients around the world on a variety of subjects such as next generation Internet networks and developing practical solutions to reduce CO2 emissions such as free broadband and dynamic charging of eVehicles. He is an author of many papers and articles on these topics and is a frequent guest speaker. For more details on my research interests see

Wednesday, July 14, 2010

Should funding agencies ban purchase of HPC clusters by university researchers?

[Many academic researchers have argued that a public cloud would never be a suitable option for many large scale HPC applications, especially due to the concerns about consistent performance. Yesterday Amazon EC2 announced that its HPC cluster service. Amazon also stated that they ran the benchmark on 880 of their Cluster Compute instances (7,040 cores) and "measured the overall performance at 41.82 teraflops using Intel's MPI and MKL libraries along with their compiler suite." These metrics place them 146 position on the TOP500 list of worldwide HPC systems! Because computational research usually involves transfer of large datasets Amazon (as well as Microsoft Azure) have acquired a number of 10G connections into various research networks. The Amazon web site lists a number of computationally intensive research projects that are already using this service – and I know that many more are in the works.

In a recent study carried out by University of British Columbia they discovered that campus HPC clusters constituted 15% of the energy consumption on that campus. As well Ed Lazowska (former chair of US President advisory committee on IT) claims in a study done at UoWahington concludes that 85% of computational research can be done using clouds (

In this age of fiscal constraint the purchase of stand alone HPC clusters for university research is making less and less sense. As well many universities are complaining about the indirect costs of research such as electricity, cooling and building maintenance. By getting rid of these very expensive HPC clusters and using clouds, universities can significantly reduce both their direct and indirect costs. These HPC facilities can also add a significant component to the university’s carbon footprint if the electricity for the campus comes from coal fired power plants. By eliminating all these HPC facilities a university can significantly reduce its carbon footprint.

What still amazes me is why neither Amazon or Microsoft have not yet developed energy and Green House Gas (GHG) reduction standards for their cloud services (e.g. ISO 14064). The savings in energy and carbon offsets could almost pay for the service entirely – and be virtually free to the academic researcher. –BSA]

Amazon Adds HPC Capability to EC2