Performance Modeling in High Performance Computing: The Convolution Method

This book provides a practical introduction to performance modeling in high performance computing and the convolution method, which is faster than traditional cycle-accurate simulation, more sophisticated than performance estimation based on system peak-performance metrics, and yields insight into the factors that affect performance and scalability on parallel computers. The author describes how to design and customize benchmark suites, event tracers, and combining operators that result in accurate tractable performance models. He also discusses cases that can improve high performance computing time-to-solution via performance modeling, including workload characterization.