Session classification for traffic aggregation

Classification is a critical task to be performed for network switches and routers to guarantee quality of service. Some applications can he categorized efficiently by using the multi-field classifiers and content-aware classifiers. However, these traditional classifications are inadequate for those applications whose classification rules can´t be configured in advance. Most of these kind applications are session based, that is more than one transport channels are needed in their communication. In this paper, we propose the state tracking session classification and present two implementation alternatives. An implantation example of session classification for H.323 VoIP is also presented in this paper. By tracking the states of the sessions, all of the packets in the session-based applications can he caught and then marked or tagged. In addition, the proposed session classification can cooperate with different aggregation techniques such as Diffserv, MPLS, and VPN to achieve efficient QoS control in a network.