Research Of QoS Queue Algorithm Based On Differentiated Services Network

With the rapid development of computer and communication technology, network size and network applications have increased with time gone by. How to meet the increasing QoS requirements of the applications is always the research focus with great practical significance. Proposed by IETF, DiffServ network model is simple and effective, scalable feature. With the principles of classifying the packets at the edge and forwarding the packets in the center, it conforms to the trend of network technology development. It is the preferred solution in the future high-speed network. As the DiffServ core technologies, queue management and scheduling are the key to network buffer management and bandwidth allocation. They have important effect on the QoS performance of the network, including throughput、Packets loss rate、Delay、Jitter. This paper proposes two improved queue algorithms on basis of DiffServ network queue algorithms. Research and innovation of this paper is as follows:1 The paper introduces QoS queuing algorithm research background and significance, then analyzes the research status both abroad and at home. We introduce basic concepts of network QoS and its indicators, study the basic concepts and network architecture about DiffServ network. Then the paper studies the usage of network simulation tools NS2 and the DiffServ network model in NS2.2 The paper focuses on the queue management algorithm in DiffServ model. Among the queue management algorithms, the paper proposes RIO-C algorithm is suitable for differentiated service model. As RIO-C discards the packets too directly, and the utilization of system for RIO-C is low, the ability to handle burst traffic for RIO-C is insufficient. We propose the INRIO-C algorithm. The relationship for average queue length and packets loss rate is nonlinear in INRIO-C. The algorithm takes the network congestion into consideration, it proposes packets loss rate method with piecewise nonlinear function, takes a smoother method for dropping packets. By NS2.35 simulation, it can protect high-priority traffic and achieve the requirements of AF queue effectively. Reducing dropping packets when network load is light, that will improve the performance of total system. Increasing dropping packets with the network load is heavy. Through network simulation and theoretical derivation, the new algorithm has been proved that it can effectively maintain the stability of the queue, reduce the packets loss rate and increase the throughput for higher priority traffic.3 The paper concentrates on the queue schedule algorithm in DiffServ network. Through studying the representative queue scheduling algorithm, and by contrast, come into the conclusion that round-robin queue scheduling algorithm is more suited to DiffServ network system. Based on WRR algorithm, which cannot solve for different packet size problem, we propose the IFWRR algorithm. The new algorithm takes the average size of the packet queue into consideration, changes in the weights reasonably, which can allocate the bandwidth more equitably. Through network simulation and theoretical derivation, IFWRR has proven that it can overcome the shortcomings of WRR in bandwidth allocation, meet the requirement for fairness between different queues according to the requirement of the actual bandwidth demand.