ROBUST SEQUENTIAL ALGORITHMS FOR TRAFFIC MONITORING IN HETEROGENEOUS DATA NETWORKS

摘要

We consider heterogeneous data networks whose traffics may be generated by either one of a number of statistically ill-specified non-parametrically defined processes and where the data generating process may change at any point in time. The objective is to effectively track the latter changes, where each acting process is essentially represented by a whole class of parametrically defined processes. We present, analyze and evaluate robust sequential algorithms which attain the objective for a variety of scenarios. Our robust algorithms consist of appropriate modifications of previously presented parametric sequential algorithms, to mainly resist the occurrence of occasional data outliers in terms of dramatic performance deterioration.