Analysis And Evaluation Of The Impact Factors On The Internet End-to-end Performance

Along with the population of the triple-play application service and the complex development of the Internet topology, it has become an urgently important issue for the ISP how to guarantee the Internet end-to-end performance. The key problem involved with the end-to-end performance comes to the effectiveness of the methodology used to analyze the possible impact factors on the end-to-end performance with quantitive evaluation of these impacts. To meet such requests, there needs improvements on both the network emulation environment and the network performance evaluation methodology. The network emulation environment should be closer to the realistic network on the emulation of the rate, packets type, and distribution of the network traffic. And the realistic models of the network equipments performance also help a lot. On the other side, the performance evaluation methodology should regulate according to different application service and reflect the dynamics on the network performance.This paper focuses on the analysis and evaluation of the impact factors on the Internet end-to-end performance from the aspects of the inter-domain routing stability and the security of end-networking. Firstly, the research of the impact of the inter-domain routing stability on the network performance includes the measurement and modeling of the router BGP processing performance used to improve the reality of the simulation, and the mathematic models describing the BGP dynamic behavior used to analyze the instantaneous stability changes on each router. Secondly, the research on the relationship between the security policies and the network performance on the end local networks mainly includes how to construct the firewall performance evaluation environment based on the realistic network traffic, and the methodology to evaluate the dynamic performance of the application service. The research in this paper employs the experimental measurements, simulations and mathematic modeling to perform the analysis and propose the relative solutions and improvements.The following lists the main work and innovational contributions in this paper:1. Here a new measurement methodology is proposed to test and model the router BGP performance and experimental models for the performance is constructed based on the data extracted form the popular commercial routers. The models describe the relationship between the router BGP performance and the configuration and the load of the router both under the normal conditions and the boundary conditions.2. Current SSFNet simulation environment is improved through the implementation of the measured empirical models describing the router BGP processing performance.3. Simulations are performed to explore routing stability including the integrated influence introduced by the multi-homing connections and the instantaneous impact on routers at different location introduced by the router BGP processing performance and memory limitation during the convergence procedure.4. New mathematic models are developed to study the BGP stability instantaneous behavior during the convergence procedure, which focuses on the instantaneous pressure and its distribution on each router with the parameters of the BGP UPDATES. Such models help a lot on the optimization of BGP routing configuration.5. A new methodology of Modeling and Generating Network Traffic (MGNT) is developed based on the classic mathematic models of the network protocols traffic and the fast analysis of the captured network peaking traffic, which can improve the measurement environment closing to the realistic network. As a initial try, this paper focuses on the TCP and HTTP traffic emulation.6. A new set of firewall performance evaluation metrics are proposed and measured in this paper, which tends to reflect the dynamics of firewall performance and end-to-end network performance on the application level. The metrics regulates according to the application protocol, such as the TCP connection setup time and the HTTP response time. The application impact factor (AIF) is used to capture the dynamics with quantitive evaluation. And the effectiveness of these metrics is proved during the measurement of current firewalls performance.