Multiresolution analysis and denoising of computer performance evaluation data with the wavelet transform

Abstract

The Wavelet Transform Analysis (WTA) offers the possibility to decompose a time series into both time and scale components. The paper applies Wavelet Transform to analyze computer performance measurement data. This type of data forms a time series that is treated and analyzed as any other signal. A noise reduction technique that detects noise components by analyzing the evolution of the Wavelet Transform modulus maxima across scales is adapted in order to improve the quality of computer performance data. This method allows to effectively remove the induced noise from measurement inaccuracies and transient random events. © 1996 Elsevier Science All rights reserved.