Department of Computer Science and Engineering

Departmental News

Button Third in a Row Best Paper Award for Systems


A team of researchers under the mentorship of Dr. Gagan Agrawal received the best paper award at the 20th International ACM Symposium on High-Performance Parallel and Distributed Computing (HPDC '11). This makes the second such recognition for Agrawal's group in the current academic year. Group members also won the best paper award at International meeting on High Performance Computing (HiPC 2010) held in December 2010. Additionally, a set of authors also were named one of three finalists for the best paper award at International Conference on Supercomputing (ICS 2011), held in June 2011. The award at this international conference finally went to another paper from OSU-CSE, specifically Dr. Zhang's team.

This most recent award at HPDC 2011 is for the paper titled: "Supporting GPU Sharing in Cloud Environments with a Transparent Runtime Consolidation Framework", authored by Vignesh Ravi, Michela Becchi, Gagan Agrawal, and Srimat Chakradhar. Vignesh Ravi is a CSE graduate student (working under Agrawal), whereas Michela Becchi and Srimat Chakradhar are from University of Missouri and NEC research, respectively.

This paper represents more of the ongoing innovations related to GPU computing from Agrawal's group (the previously recognized papers from HiPC 2010 and ICS 2011 were also related to GPUs). GPUs or Graphics Processing Units were originally designed for games and other graphics applications, but in the last 3-4 years, have been identified to be suitable for general purpose computations, essentially providing "supercomputing" at a very cheap price. This specific work from Agrawal's group takes GPUs even a step further, as each GPU is now shown capable of scaling multiple general purpose computations at the same time. A software framework and mechanisms have been developed, which allow programs from independent users to transparently share a single GPU. The net result is significant advantage in cost-effectiveness and throughput of the device.

The HPDC is the premier computer science conference for presenting new results relating to large scale high performance parallel and distributed systems used in both science and industry. In 2011, HPDC received more than 170 submissions. The program committee selected less than 15% of all submissions for presentation at the conference, and during their deliberations, identified the paper from Ohio State as the strongest submission.

OSU logo