95. Marrying HPC and Cloud for Long Term Happiness
Authors: Apoorve Mohan (Northeastern University)Ravi S. Gudimetla (Northeastern University)Ata Turk (Boston University)Sourabh Bollapragada (Northeastern University)Rajul Kumar (Northeastern University)Jason Hennessey (Boston University)Evan Weinberg (Boston University)Dimitri Makrigiorgos (Boston University)Christopher N. Hill (Massachusetts Institute of Technology)Gene Cooperman (Northeastern University)Peter Desnoyers (Northeastern University)Richard Brower (Boston University)Orran Krieger (Boston University)
Abstract: Traditional HPC clusters, with deep job submission queues, by construction are always almost fully utilized. On the other hand, the cloud has time-varying workloads, and the cloud business model depends on being underutilized to instantly support all customer requests. A marriage of these two environments could provide additional resources to HPC users while offering increased utilization to the cloud. In this poster, we present the frameworks we built to enable a successful symbiotic co-existence of HPC and the cloud and showcase the benefits achievable with a prototype deployment.
Two-page extended abstract: pdf