Path And Link Loss Inference In IP Network Based On Active Probing

With the rapid development of the Internet, people dependent on network service level continuously improve, and they also put forward higher requirements on the performance of the network. Network performance measurement is the basis and verification means of analysis of network traffic, topology, behavioral modeling. How to use low cost and accurate measuring for the operational status of the entire network become the research hotspot to researchers. The network link loss rate can reflect the operational status and the congestion level of the network in the process of IP network’s transmitting information, it is an important indicator of network performance, so how to quickly and accurately get link loss rate is becoming a hot spot of many researchers. This paper focuses on the inference of link loss rate.In this paper, we propose an end-to-end path loss inference algorithm with network tomography. We can infer the loss rate of the remaining paths accurately by measuring r independent paths that with uniform distribution. The new method adaptively selects the next path by using the results of a probe in the previous detection returned. The deletion of the non-congested paths and links makes the size of routing matrix smaller, which reduces the number of independent paths obviously. And our method can obtain the loss rates of all remaining paths without solving the value of the link loss rate, which greatly reducing the complexity of the algorithm. The simulation results show that our method not only guarantee the accuracy of inference, but also decrease the number of independent paths to5%～14%compared to the existing classical method.In order to address the high computational complexity, the limited application and difficult to detect defects of the existing algorithms, we propose a novel link loss rate inference algorithm. We first establish algebraic model, obtain the loss rate of determined links, and remove all these determined links from the system, then divide the links in the network topology into several independent subsets, and finally converted the solution of the equations of each subset to optimization problem of non-linear programming, we can get more accurate results with a smaller price. Through software simulation, we compared our method with LIA algorithm and LIABLI algorithm, our method only uses2%of LIA’s probe cost, but the diagnostic accuracy, inference accuracy and running time was significantly improved.