Adaptive Performance Modeling on Hierarchical Grid Computing Environments

In the past, efficient parallel algorithms have always been developed specifically for the successive generations of parallel systems (vector machines, shared-memory ma-chines, distributed-memory machines, etc.). Today, due to many reasons, such as the inherent heterogeneity, the di-versity, and the continuous evolution of the existing parallel execution supports, it is very hard to solve efficiently a tar-get problem by using a single algorithm or to write portable programs that perform well on any computational supports. Toward this goal, we propose a generic framework based on communication models and adaptive approaches in order to adaptively model performances on grid computing environ-ments. We apply this methodology on collective communi-cation operations and show, by achieving experiments on a real platform, that the framework provides significant per-formances while determining the best combination model-algorithm depending on the problem and architecture pa-rameters