MPOA flow classification design and analysis

We propose a framework for the performance analysis and flow classification design of multi-protocol over ATM (MPOA) network. The study based on the real Internet/intranet traces shows that even at high cost with long delay for each shortcut setup, MPOA can offer significant performance gain over the traditional routed network in an inter ELAN communication environment. In comparison, the MPOA performance gain in an Internet backbone environment is much less significant, mainly because of the dominant short-lived flows contributed by both d.n.s. and h.t.t.p. applications. We also propose a flow classification algorithm, which substantially reduces the implementation complexity while achieves the same level of performance as compared to the default flow classification algorithm proposed by MPOA standard. A simple timeout mechanism is also introduced to the flow cache table management for significant performance improvement. We further develop a stable, adaptive flow classification algorithm, which achieves a near-optimal solution to minimize the constrained MPOA resource utilizations.