Manila: Using a densely populated PMC-space for power modelling within large-scale systems.

Highlights • Explores the importance of coverage and density of a PMC-space for power modelling. • Shows a k-nearest neighbour search can be used to estimate power from the PMC-space. • Discusses the simplicity and usability of the Manila modelling methodology. • Demonstrates our model is portable across system architectures. • Examines the scalability of monitoring cluster power. Abstract In this paper, we propose a new power modelling method, called Manila, that can largely reduce the effort of PMC-based power modelling using multi-dimensional k-nearest neighbour searching, without the use of model tuning and domain-specific knowledge. This method helps improve the accuracy of PMC-based power modelling and widen its scope of use. Specifically, Manila uses a densely populated PMC-space, generated from a large variety of computing workloads, which is essential for our improvement. This is in contrast to current PMC-based power models, that have a sparse PMC-space, due to using a predefined benchmark suite. A parameterised micro-benchmark is used to bootstrap the PMC-space, which is independent from any applications and can generate the generic computing workloads of many applications. Our method is more widely extendable and applicable than the existing methods. Manila can efficiently search the dense PMC-space for power estimates using a nearest neighbour search algorithm. Experimental results demonstrate that Manila is more accurate in power measurements for a wide range of parallel benchmarks, with an average mean absolute error of 2.8%. [ABSTRACT FROM AUTHOR]