An adaptive flow classification algorithm for IP switching

Flow classification is one of the key issues in IP switching. To achieve better performance, flow classification should be matched to the varying IP traffic and an IP switch should make use of its hardware switching resources as fully as possible. Based on these arguments, this paper proposes an adaptive flow classification algorithm for IP switching. By dynamically adjusting the values of its control parameters in response to the present usage of the hardware switching resources, this adaptive algorithm can efficiently match the varying IP traffic and thus improve the performance of an IP switch. A simulation study based on real IP traces is performed to demonstrate the efficiency of this adaptive algorithm and to show the improvement of the system performance.