Virginia Tech: A Major Presence at IEEE IPDPS 2009

Publish Date: 07/24/2009

Virginia Tech, an emerging leader in high-end computing research, had a substantial presence at last month's IEEE International Parallel & Distributed Processing Symposium (IEEE IPDPS), held May 25-29, 2009, in Rome, Italy. Highlighting Virginia Tech's presence are four accepted papers in the main IPDPS conference, three papers in associated IPDPS workshops (including one on "The Green500 List"), and a keynote address by Dr. Wu Feng at the IPDPS Workshop on Communication Architecture for Clusters.

With a conference acceptance rate of only 23%, Virginia Tech was a significant presence with four papers featuring research experts from the Department of Computer Science. The presented papers were:

• “CellMR: A Framework for Supporting MapReduce on Asymmetric Cell-Based Clusters” (Rafique, Rose, Butt and Nikolopoulos). PhD candidate Muhammad Mustafa Rafique presented this paper. According to Rafique, this research serves as a “step towards proving the necessary software underpinnings to enable next generation HPC applications to utilize emerging hardware such as IBM Cell processors and graphics processors.”

• “Elastic Scaling of Data Parallel Operators in Stream Processing” (Schneider, Andrade, Gedik, Biem and Wu). PhD candidate Scott Schneider described work that he has done in collaboration with researchers at IBM. The goal is to dynamically adapt the degree of parallelism used for streaming operations, in response to both external load and changing resources.

• “Scheduling Resizable Parallel Applications” (Sudarsan and Ribbens). PhD candidate Rajesh Sudarsan described his work in run-time systems and middle-ware for more effective utilization of large parallel clusters. The experimental work reported in this paper made extensive use of Virginia Tech’s new System G supercomputer.

• “Multi-Dimensional Characterization of Temporal Data Mining on Graphics Processors” (Archuleta, Cao, Feng, Scogland). PhD Candidates Jeremy presented a paper on the potential of the graphics-processing unit (GPU) to vastly accelerate discovery and innovation across a multitude of disciplines. This research presented a characterization of a MapReduce-based data-mining application on a general-purpose GPU.

Presentations and Workshops: Dr. Feng presented a keynote address at the IPDPS Workshop on Communication Architecture for Clusters and served as co-chair of the IPDPS Workshop on High-Performance, Power-Aware Computing.

Three papers featuring VT research were presented in the associated workshops including “The Green500 List” by Dr. Feng.