Skip to Main Content
Ordinal time series analysis is a simple approach to the investigation of experimental data. The basic idea is to consider the order relations between the values of a time series and not the values themselves. This results in fast and robust algorithms that extract the basic intrinsic structure of the given data series. This paper gives a short overview of this approach and describes the application of simple ordinal time series methods like rank autocorrelation, local rank autocorrelation and permutation entropy to workload traces from large grid computing systems. We show how these methods can be used to extract important correlation information from experimental traces and how these methods outperform traditional methods.