Thursday, July 16

MS52
Graph Partitioning and Dynamic Load Balancing (Part I of II)

10:30 AM-12:30 PM
Room: Sidney Smith 1085

Graph partitioning has extensive applications in many areas, including scientific computing, task scheduling, VLSI design, and data mining. Significant progress has been made in developing fast and high quality graph partitioners. Focus has now shifted to efficient methods for partitioning and load balancing in dynamic and adaptive settings. Dynamic load balancing not only requires a rapid partitioner, but also an intelligent mapping of partitions to processors to reduce the data redistribution cost. This is important for problems with time-varying workloads and parallel systems with non-dedicated resources. The speakers will describe several partitioning and load-balancing algorithms, and results from recent applications to realistic problems.

See Part II, MS66.

10:30 Dynamic Load Balancing for Problems with Geometric Locality

Bruce Hendrickson, Sandia National Laboratories, Albuquerque

11:00 An Adaptive and Parallel Computational Framework for Partial Differential Equations

Joseph Flaherty, R. Loy, J. Teresco, M. S. Shephard, and L. Ziantz, Rensselaer Polytechnic Institute

11:30 Graph Partitioning with Constraints

Alex Pothen, Old Dominion University, and ICASE- NASA Langley Research Center

12:00 Multi-Objective and Multi-Constraint Graph Partitioning

Kirk Schloegel, University of Minnesota, Minneapolis; Vipin Kumar, Organizer; and George Karypis, University of Minnesota, Minneapolis