A NOVEL REAL-TIME SELF-SIMILAR TRAFFIC DETECTOR/FILTER TO IMPROVE THE RELIABILITY OF A TCP BASED END-TO-END CLIENT/SERVER INTERACTION PATH FOR SHORTER ROUNDTRIP TIME

摘要

The self-similarity (S{sup}2) filter is proposed for real-time applications. It can be used independently or as an extra component for the enhanced RTPD (real-time traffic pattern detector) or E-RTPD. The S{sup}2 filter basis is the "asymptotically second-order self-similarity" concept (alternatively called statistical 2nd OSS or S2nd OSS ) for stationary time series. The focus is the IAT (inter-arrival times) traffic. The filter is original because similar approaches are not found in the literature for detecting self-similar traffic patterns on the fly. Different experiments confirm that with help form the S{sup}2 filter the FLC (Fuzzy Logic Controller) dynamic buffer size tuner control more accurately. As a result the FLC improves the reliability of the client/server interaction path leading to shorter roundtrip time (RTT).