摘要
对共享缓存分组交换机提出了按需分配缓存的具有门限的动态双队列缓存管理策略。具有门限的动态双队列特性,在提高不同优先级业务服务公平性以提供服务质量保证的同时,也提高了缓存的整体有效利用率。据此对该系统建立了M1+M2/M/1/K+(K)的排队模型,给出了相应的Q矩阵,利用矩阵几何解方法对模型求解得到了稳态概率分布。最后做了相应的性能分析,定量给出了不同优先级分组的队长分布、丢失概率等的计算公式。
A buffer management scheme of dynamic dual queue with queue threshold (DDQQT) is proposed for a shared buffer packet switches, the scheme can allocate buffer on demand. The key of the dynamic dual queue with queue threshold is to enhance the fairness to meet the QoS (Quality of Service) guarantees for different service classes and the overall utilization of the buffer. An associated M1-b-M2/M/1/K+ (K) queueing model is established, Q matrix is given, stead-state probability distribution is worked out by using matrix-geometric solutions. Based on it, some performance evaluations such as queue length densities and loss probabilities of the two classes packets are given.
出处
《计算机科学》
CSCD
北大核心
2008年第1期60-63,93,共5页
Computer Science
基金
国家自然科学基金资助项目:互联网通信排队模型研究(70571030)
马尔可夫随机场及非齐次马氏链的极限定理(10571076)
关键词
分组交换
服务质量
缓存管理
排队模型
动态双队列
Packet switching, Quality of service, Buffer management, Queueing model, Dynamic dual queue