Scalable Policing Using Statistical Pattern Recognition

In this report, a framework for the Policing function in the Scalable Resource Reservation Protocol (SRP) is proposed. SRP is characterized by a high scalability, soft guarantees and no a priori specifications of the traffic profile. Resources are reserved for the whole set of elementary stream, the aggregate flow. The policing function monitors the aggregate flow and performs a conformance test to detect packets that belong to conformant or non-conformant flows. If the packets in the aggregate flow exceed the expected number, the policing function has to make a selection to discard some of them. We adopt a Statistical Pattern Recognition (SPR) for the packet classification. Aim of the policing function is to limit the impact of non-conformant traffic on conformant traffic, and also to reduce usefulness of the former. We first describe the approach, then we trace a theoretical analysis and, finally, we describe the simulation carried out and the experimental results.