Integrating algorithmic and systemic load balancing strategies in parallel scientific applications

Load imbalance is a major source of performance degradation in parallel scientific applications. Load balancing increases performance of parallel applications in distributed environments. At a coarse level of granularity, advances in runtime systems have been proposed in order to control available resources using task migration. At a finer granularity level, advances in algorithmic strategies for dynamically balancing loads by data redistribution have been proposed. Algorithmic and systemic load balancing strategies have complementary set of advantages. An integration of these two techniques should result in a system, which delivers advantages over each technique used in isolation. This thesis presents a design and implementation of a system that combines an algorithmic load balancing strategy called Fractiling with a systemic load balancing system called Hector. It also reports on experimental results of running N-body simulations under this integrated system. The experimental results indicate that the integrated system provides performance improvement for large applications.