On utilizing experiment data repository for performance analysis of parallel applications

Performance data usually must be archived for various performance analysis and optimization tasks such as multi-experiment analysis, performance comparison, automated performance diagnosis. However, little effort has been done to employ data repositories to organize and store performance data. This lack of systematic organization of data has hindered several aspects of performance analysis tools such as performance comparison, performance data sharing and tools integration. In this paper we describe our approach to exploit a relational-based experiment data repository in SCALEA which is a performance instrumentation, measurement, analysis and visualization tool for parallel programs. We present the design and use of SCALEA's experiment data repository which is employed to store information about performance experiments including application, source code, machine information and performance data. Performance results are associated with experiments, source code and machine information. SCALEA is able to offer search and filter capabilities, to support multi-experiment analysis as well as to provide well-defined interfaces for accessing the data repository and leveraging the performance data sharing and tools integration.