Analysis of fGn and http Requests Traces Using Localized Multiscale H Parameter Estimation

Abstract

For effective employment of the traffic models in real applications, the ability to detect singularities embedded in the signal is very important. Events like weekly pattern variations, equipment malfunctions, or holiday periods affect signal characteristics, making it difficult to obtain precise traffic parameters. By performing analysis locally over different scales, a map of estimated parameters can be computed. This article presents a concept of time-scale estimated H parameter decomposition (called H surface) and methods applied to detect variations in the traffic characteristics. Results based on synthetic fGn signals and http requests traces are presented.