|
|||
|
IMADA - Department of Mathematics and Computer Science |
|
Performance and workload modeling has numerous uses at every stage of the high-end computing lifecycle: design, integration, procurement, installation, tuning, and maintenance. Despite the tremendous usefulness of performance models, their construction remains largely a manual, complex, and time-consuming exercise. We propose a new approach to the model construction, called modeling assertions (MA), which borrows advantages from both the empirical and analytical modeling techniques. Modeling assertions generate symbolic performance models of a target application. This strategy has many advantages over traditional methods: incremental construction of a realistic performance model, straightforward model validation against empirical data, and intuitive error bounding on individual model terms. We demonstrate this new technique on the NAS parallel CG and SP benchmarks by constructing high fidelity models for the floating-point operation cost, memory requirements and MPI message volume. These models are driven by a small number of key input parameters thereby allowing a user to efficiently explore the design space for future architectures. Our evaluation shows that MA does make the construction and use of symbolic performance models more practical than either empirical or analytical modeling techniques alone. System independent performance models provide additional advantages particularly on unique systems architecture. As a case study, we conduct in-depth performance analysis of grand challenge molecular dynamics simulations on recent parallel systems, Blue Gene/L and XT3, and on non-conventional processing architectures including an FPGA-based system and a multi-threaded architecture. We identify the limitations of traditional performance modeling approaches and conclude that the MA performance models enable us to explore an optimal workload mapping on these diverse system architectures. Host: Peter Kornerup SDU HOME | IMADA HOME | Previous Page Last modified: Thu Dec 15 11:22:36 CET 2005 Joan Boyar (joan@imada.sdu.dk) |