一种支持区分服务的模糊公平分组丢弃算法

随机早期检测(RED)算法是目前Internet中广泛使用的一种缓存管理算法，对RED算法进行改进以提高公平性以及对网络动态变化的适应性一直是主动队列管理的研究重点之一。该文引入新的变量分享指数，取代单个业务流缓存占用量来表征不同业务流对网络资源的占用情况，以克服缓存占用量表征公平性的缺陷。同时，在无线网络中对分享指数的定义进行推广，使其同时兼顾公平性和系统性能(信道条件)。通过对分享指数和队列长度共同制定规则，提出了一种基于模糊逻辑控制的缓存管理算法。算法在根据缓存队列长度计算丢弃概率时，根据不同流对网络资源的占用情况(即分享指数)进一步决定增加或减小相应的丢弃概率，从而可利用不同流之间的差异以加速缓解拥塞，具有更好的自适应性和公平性。仿真表明，采用该算法后，系统性能得到改善。

A Fuzzy Fair Packet Dropping Algorithm Supporting Differentiated Services

Random Early Detection (RED) is a kind of buffer management algorithms which is widely used in the current Internet. To improve the adaptability and fairness of RED is one of the main tasks of the active queue management. In this paper, the buffer occupancy of individual flow is replaced by a new introduced variable - Sharing Index to indicate the network resource share among different flows, so as to overcome the limitation of buffer occupancy. The definition of Sharing Index in wireless networks is also extended to make it consider both fairness and system performance (channel condition). Through defining rules of both Sharing Index and queue length, a fuzzy logic based buffer management algorithm named FF-RED is proposed. When calculating packet dropping probability in FF-RED algorithm, the probability will further increase or decrease based on the sharing of network resource (by means of Sharing Index), so it can use the difference between flows to make congestion relief more quickly, and it has better adaptability and fairness. Simulation results show that the system performance is improved through using the proposed algorithm.