Performance Analysis of KDD Applications using Hardware Event Counters

Abstract

Modern processors and computer systems are designed to be efficient and achieve high performance with applications that have regular memory access patterns. For example, dense linear algebra routines can be implemented to achieve near peak performance. While such routines have traditionally formed the core of many scientific and engineering applications, commercial workloads like database and web servers, or decision support systems (data warehouses and data mining) are one of the fastest growing segments in the high-performance computing market. Many of these commercial applications are characterised by complex codes and irregular memory access patterns, which often result in a decreased performance. Due to their complexity and the lack of source code, performance analysis of commercial applications is not an easy task. Hardware performance counters allow acquisition of low level, reliable data, necessary to perform detailed analysis of program behaviour. In this paper we describe experiments and present first results conducted with various KDD applications on an UltraSPARC III platform.