Discovering Periodic Patterns in System Logs
Marcin Zimniak, Janusz R. Getta, Wolfgang Benn · 2014
Abstract. Historical information stored in the software system logs, au-dit trails, traces of user applications, etc. can be analysed to discover the patterns in periodic variations of levels and structures of the past work-loads. These patterns allow for the estimation of intensities and struc-tures of the future workloads. The correctly anticipated future workloads are used to improve performance of software systems through appropriate allocation of computing resources and through restructuring of associ-ated system support. This work defines a concept of periodic pattern and presents the algorithms that find the periodic patterns in the traces of elementary and complex operations on data recorded in the system logs. 1