摘要
为了求解以长程相关业务作为输入的网络端到端延迟上界,提出了基于分形漏桶和GPS系统的端到端网络模型,利用网络演算理论证明了分形漏桶的引入不增加网络端到端延迟上界,建立了端到端延迟理想及其近似确定上界模型。数值结果表明,分形漏桶有效改善了长程相关特性对端到端延迟上界所造成的影响,与简单漏桶相比,基于分形漏桶的长程相关业务端到端延迟上界具有更好的紧致性。
In order to get upper bounds on end-to-end delay of networks with LRD （long-range dependence） input traffic, a network model was presented based on FLB （fractal leaky bucket） and GPS （generalized processor sharing）, and the result that FLB do not increase the bounds of network delay was proved by network calculus, alld the optimal and asymptotic deterministic upper bounds of end-to-end delay of LRD business were derived. Numerical results demonstrate that FLB can improve the effect of the self-similar degree of LRD business on the end-to-end delay, and bounds on end-to-end delay based on FLB are tighter than those based on leaky bucket.
出处
《通信学报》
EI
CSCD
北大核心
2008年第7期32-38,43,共8页
Journal on Communications
基金
国家自然科学基金资助项目(60773012,60573127)
中国博士后科学基金资助项目(20070420782)
湖南省自然科学基金资助项目(07JJ612T)~~
关键词
网络演算
通用处理器共享系统
分形漏桶
端到端延迟
network calculus
generalized processor sharing
fractal leaky bucket
end-to-end delay
