Guest Speaker

Message Passing for a Million Processes

Pavan Balaji
Argonne National Laboratory
Argonne National Laboratory

May 15 2009 12:30PM
480 Dreese Labs
All interested parties are welcome.
Pizza lunch will be provided.

Abstract:
Upcoming exascale capable systems are expected to comprise more than a million processing elements. Although message passing (in the form of MPI) is the dominant form of parallel programming today, mainly because of its capability to explicitly manage data and optimize data locality, some researchers and users wonder whether MPI will scale to such large processor counts. In this talk, I will focus on a subset of issues that need to be addressed as we think about allowing MPI to work efficiently on exascale platforms. These include scalability issues with the MPI standard as well as current implementations,process management issues related to scalably managing a million processes, and MPI's ability to function efficiently in hybrid environments involving MPI + threads or MPI + Partitioned Global Address Space models.

Bio:
Dr. Pavan Balaji holds a joint appointment as an Assistant Computer Scientist at the Argonne National Laboratory and as a research fellow of the Computation Institute at the University of Chicago. He had received his Ph.D. from the Computer Science and Engineering department at the Ohio State University. He has nearly 60 publications in various sub-areas of high performance computing and has delivered more than 60 talks and tutorials at different conferences and research institutes. He has received several awards for his research activities including the Director's Technical Achievement award at Los Alamos National Laboratory, and many best paper and other awards. Dr. Balaji has also served as a Program Co-chair or Co-editor for more than half-a-dozen conferences, workshops and journals, and has been on the TPC for numerous International conferences.

Host: Dhabaleswar Panda