Research On Queue Buffer Management Based On Traffic Prediction In DiffServ Architecture

With the development of network technology, the requirement of the real-time and multimedia application based on network is increasing. The guarantee of QoS(Quality of Service) has become one of the research hotspots on network. Differentiated services model put forword by IETF provides an effective way to solve the QoS problem. This paper is to study the queue buffer management in DiffServ architecture. The queue buffer management plays a significant role in the network transmission control, and it is one of core technologies used in QoS control.DiffServ model provides different QoS services based on aggregated flows. The basic idea is to classify the traffic according to QoS requirements on the network edge, and different class realizes different forwarding characteristic on the core router. It is characterized by the use of the aggregated mechanism, it has the same characteristic traffic get together, and provides services for the aggregated flows, so it has the good scalability.Queue buffer management is to manage buffer resources in the network transmission node. In the process of packet transmission, the network node which packet flows through usually uses the queue buffer, delay forwarding to improve the bandwidth utilization of output link. When packets arrive at the queue, the queue buffer management mechanism decides whether to allow the packet to enter into the buffer queue according to certain information and strategy. In this paper, a new queue buffer management algorithm--the adaptive RIO-C based on traffic prediction algorithm is proposed. The algorithm combines RIO-C algorithm used in DiffServ networks with ARED algorithm, and applies RLS algorithm to predictive traffic. It aims at improving the EF service's performance in the DiffServ network. In order to verify the validity of the new algorithm, two simulation experiments are carried on under the NS2 network simulation environment:one for ftp traffic transmission, the other for ftp and cbr traffic transmission. The results showed that the new algorithm could reduce the latency and packet loss rate, and could stable latency jitter effectively, thus enhanced the QoS performance.