MPOA flow classification design and analysis based on neura1 network technique

摘要

In this paper, we develop a framework for the flow classification (FC) design and performance analysis of multi-protocol over ATM (MPOA) network. We propose an FC algorithm, which substantially reduces the implementation complexity while achieving the same level of performance as compared to the default FC algorithm proposed by MPOA standard. We further design an adaptive, self learning system to achieve a near-optimal flow cache table management in terms of performance gain. The self learning system is then used for the performance analysis of MPOA. The simulation study based on the real Internet/Intranet traces shows that MPOA can offer significant performance gain in an inter-ELAN environment.