For many decades, if you belonged to the “chosen people,” supercomputers were on the floor of your HPC center – mighty, powerful, thundering, admired, the Holy Grail of science and engineering. And you were one of the few elected HPC-savvy; you were the magician conjuring grand challenge solutions out of this magic box on the floor.
Over the last decade, however, we saw some of this HPC moving off the floor. Multicore CPUs and GPGPUs, advanced memory hierarchies, and great graphics came to the desk of the engineer and the scientist, a mini-Cray to hug, known as a workstation by the unromantic…
This whole HPC spectrum, from the lone wolf supercomputer to the winged HPC in the cloud, can be experienced at the ISC Conference in Leipzig. And some of the most representative sessions of HPC in the cloud (or HPC as a Service) are of course the Monday session on HPC Cloud – From Algorithms to Applications, and the Wednesday session on HPC – From Desktop to Cloud.
But these sessions are, in fact, just the overture for the real thing: ISC Cloud’13 this September in Heidelberg. This even encapsulates a rich program with talks from over 35 top experts from HPC and digital manufacturing with actual hands-on experience in HPC and engineering in the cloud. You will find a lot insight in the clouds of Heidelberg, but be careful; you might also lose your heart in this beautiful city.
Why do we believe the window to HPC in the cloud is now wide open? One proof point: At the Simulia Community Conference in Vienna two weeks ago, I asked 30 engineers about using HPC for engineering simulations. The result: 17 engineers were (mainly) using workstations and 13 were using HPC clusters, but with 21 out of the 30 indeed interested in cloud computing for engineering simulations.
Two years ago, at the European COMSOL Multiphysics Conference, the result was completely different. From the 52 engineers I interviewed, 50 were using workstations, and only 2 were using HPC clusters. Nobody was interested in cloud computing for engineering applications back then.
A more recent proof point is the UberCloud HPC & CAE Experiment, which started in August last year. This experiment explored the end-to-end process of accessing and using HPC clouds for digital manufacturing in areas such as CFD, CAE, bio, and other areas. Teams of four (end-user, ISV, hardware providers, and HPC expert) were used to support each application and after just 10 months, 81 teams have been formed with almost 500 organizations participating from 46 countries.
This trend is leading us to new realm, where HPC, available on demand, will become mainstream, and will be able to conquer the ‘missing middle’.
This blog was originally published on ISC Events and is reproduced here with permission.