VERIFYING LARGE-SCALE SYSTEM PERFORMANCE DURING INSTALLATION USING MODELING.pdf

VERIFYING LARGE-SCALE SYSTEM PERFORMANCE DURING INSTALLATION USING MODELING Darren J. Kerbyson Los Alamos National Laboratory Performance and Architectures Laboratory (PAL), CCS-3 P.O. Box 1663, Los Alamos, NM 87545 djk@ Adolfy Hoisie Los Alamos National Laboratory Performance and Architectures Laboratory (PAL), CCS-3 P.O. Box 1663, Los Alamos, NM 87545 hoisie@ Harvey J. Wasserman Los Alamos National Laboratory Performance and Architectures Laboratory (PAL), CCS-3 P.O. Box 1663, Los Alamos, NM 87545 hjw@ Abstract In this paper we describe an important use of predictive application performance modeling - the validation of measured performance during a new large-scale system installation. Using a previously-developed and validated performance model for SAGE, a multidimensional, 3D, multi-material hydrodynamics code with adaptive mesh refinement, we were able to help guide the stabilization of the Los Alamos ASCI Q supercomputer. This system was installed in several stages and has a peak processing rate of 20-Teraflops. We review the salient features of an analytical model for SAGE that has been applied to predict its performance on a large class of Tera-scale parallel systems. We describe the methodology applied during system installation and upgrades to establish a baseline for the achievable “real” performance of the system. We also show the effect on overall application performance of certain key subsystems such as PCI bus speed and processor speed. We show that utilization of predictive performance models can be a powerful system debugging tool. Keywords: Performance Modeling; Terascale Systems; Performance Validation; High Per- formance Computing. 21. Introduction Performance modeling is a key approach that can provide information on the expected performance of a workload given a certain architectural configuration. It is useful throughout a system’s lifecycle: starting with a design when no hardware is available for measurement, in procurement for the comparison