The Reliability Modeling And Analysis Of Self-similar Network

Measurements and research appear that computer networks show the self-similar traffic feature (self-similar property of network). The property is valuable to network performance analysis and network design. Network reliability is one of the important factors of its performance analysis and design. Reliability analysis based on self-similar network is becoming one of the hotspot of network analysis. And the reliability of the self-similar network is mainly researched in this thesis. Network traffic performance analysis models are presented on the base of traffic models. Based on these models, reliability models are presented further. And then, the analysis and simulation works is done about them.The related work of reliability is introduced firstly. And the availability index is chosen to be the measurement of network reliability. The common self-similar traffic models which are used to describe the traffic processes are also presented. Then ON/OFF model imposing on mass ON/OFF sources to generate self-similar traffic is chosen to be the traffic model.Based on ON/OFF model, system composed of network node and traffic was viewed as a G/M/1/B queuing model. The effective bandwidth model of self-similar network is built for the bandwidth distribution of ON/OFF sources. The stationary queue length distribution was derived by large deviation theory firstly. Then the effective bandwidth is deduced by combining with overflow probability. An average queuing delay (AQD) analysis model for self-similar network is presented for the queuing delay. Stationary average queue length was got by solving the average value of the distribution. Then the average queuing delay time is derived by Little's theory. Furthermore, an average packets loss rate (APLR) model for self-similar network is presented for the network packets loss. Stationary packets loss rate distribution is derived according to the relation between queue length and packets loss rate. Then average packets loss rate model is got by solving the average value of packets loss rate. The simulation work for AQD and APLR is completed by NS-2 to verify the validity of the analytical results.The formulization of the network reliability index is completed based on traffic performance. According to AQD model and APLR model, two reliability models based on single valve are built respectively: reliability based on AQD and reliability based on APLR. Analysis work about the two models is also done subsequently. Then reliability vector model is implemented in order to measure reliability more comprehensively.Self-similar network reliability simulator (SNRS) is designed and implemented lastly based on the analysis to the disadvantages of present network simulators. SNRS could be used to model AQD and APLR of self-similar network, then to analyze reliability. By experiments, the simulation results are tally with analysis conclusions, so the validity of the reliability models above is verified.