Dynamic Data Distribution and Processor Repartitioning for Irregularly Structured Computation

Irregular applications comprise a significant and increasing portion of jobs running in parallel environments. Recent research has also shown that both system utilization and application throughput improves in parallel environments when resources allocated to applications can be
reconfigured dynamically at run-time. We have implemented the Distributed Resource Management System (DRMS) that supports the development and execution of regular and irregular reconfigurable
applications in time-varying resource environments. In this paper, we discuss DRMS support for writing reconfigurable irregular applications and describe the dynamic data redistribution mechanisms, in some detail. We also present performance levels achieved by the data
redistribution primitives, using a sparse Cholesky factorization algorithm as a model irregular application. Our results show that the cost of dynamic data redistribution among different processor
configurations for irregular data are comparable to those for regular data.