The Research And Simulation Of Mechanism For Multi-priority Queues Management

With the development of the Internet, the guaranteeing of Quality of Service (QoS) has become an important problem. Active queues management and packet scheduling algorithms are important contents of QoS.Firstly, this thesis analyzes the recent queues management algorithms: RED(random early detection) and its derivative algorithms. Then by introducing 2max_th parameters that are used in RED_gentle, an improved queue management algorithm on the basis of WRED(weighted random early detection) :WRED_gentle, is proposed. This new algorithm can achieve Diffserv and slightly weaken the sensitivtity of max_p parameter and enhance the robustness of WRED algorithm.Secondly, this thesis analyzes the recent packet scheduling algorithms: PQ(priority queueing), EDF(earliest deadlined first) algorithms and so on. An advanced dynamic priority scheduling algorithm combining slot and priorities: PQBEDF(priority queue based on EDF) algorithm, is analyzed. Because the dynamic priority is changing too quickly and the Quality of Service for higher priority queues is reduced, an improved packet scheduling algorithm: PQBEDF⁺ is presented. Depending on the actual situation of used buffer, PQBEDF⁺ algorithm introduces threshold for different businesses and enhances its fairness guarantee. At the same time, buffer management algorithms are also considered. Since the drop-tail or drop-head discarding strategy is implemented in PQBEDF algorithm, the concept of packet priority does not exist and Diffserv can't be provided. Hence PQBEDF⁺ algorithm and WRED_gentle algorithm are combined to achieve Diffserv and guarantee the end to end delay of packets, which obviously increases the effectiveness of network resource allocation.Finally, the WRED_gentle and PQBEDF⁺ algorithm described above are simulated and proved with OPNET software. Experimental results show that the performance of algorithms are improved greatly.