Time Series Metric Extraction for Application Behaviour Modeling
Riccardo Maganza · Zenodo (CERN European Organization for Nuclear Research) · 2020
Starting from univariate signals coming from test workloads run under different restriction a consistent single-dimensional signal parametrization method through piecewise-linear functions was defined and packaged in a ready-to-use suite. The results could be used to characterize the signals and measure their reactions to changes in computing and/or network resources. More work is still needed for the multidimensional case.