Purdue University

The Rosen Center for Advanced Computing (RCAC) at Purdue University has opened up opportunistic access to 11 teraflops of computing power to the TeraGrid community. Based on a new model known as community clusters developed by researchers at RCAC, this new computing resource will be accessible to TeraGrid researchers and educators using Condor. The community cluster currently supported by RCAC includes 1024 Xeon 64-bit (Irwindale) processors, 194 Opteron 64-bit processors with InfiniBand interconnects, and 618 Xeon 32-bit processors – a combined capacity of 11 TFlops.

Community clustering is an innovative idea gaining significant momentum in the national and international cyberinfrastructure community. Under this model, researchers pool funds to contribute cluster nodes that are centrally managed by RCAC. In return, researchers get priority access on the machines they purchased as well as preemptable access to any unused cluster nodes owned by other researchers. The community clusters have dedicated queues for each resource owner as well as a preemptive queue that can access all the computational power of the cluster even nodes that researchers don’t own. Backfilled by Condor the unused cycles are made available to the general user community. A Condor job manager has been installed on the Purdue TeraGrid Globus gatekeeper to route TeraGrid jobs to these clusters. The most simple way to access them is to submit a Condor-G job that points to tg-gatekeeper.purdue.teragrid.org/jobmanager-condor.

“Faculty members are increasingly receiving funding through grants or startup packages to deploy their own supercomputers. These dedicated resources can be bundled together to create significant computational power. But these resources are not used all the time and a resource sharing mechanism needs to be defined to allow access to other local or national users. We have taken a first step towards true resource sharing and grid computing by deploying a condor job manager for TeraGrid users. Even though no cycles are guaranteed, we believe that significant scientific discovery and learning can be enabled through the use of our community clusters,” said Dr. Sebastien Goasguen, Senior Research Scientist and the TeraGrid Site Lead at Purdue University.

Dr. Gerhard Klimeck, the Technical Director of the NSF-funded Network for Computational Nanotechnology noted, “For my research in computational nanotechnology I need dedicated access to computing power. However, when my students or I are not using the machine I am glad to share it with anyone. The community cluster model put in place by RCAC is excellent and I have an incentive to invest in it. Giving back to the TeraGrid community through my unused cycles is a very obvious and practical thing to do."

Non-Purdue researchers can request access to this new resource through the POPS system at rcac-info@purdue.edu.