Research On Multi Metric Network Topology Estimation Method

With increasing network size and complexity, in order to network performance timely and effective detection and control, real-time access to accurate network topology has a very important significance. In the case of internal nodes does not require network collaboration, access path state parameter measuring method by measuring end topology estimation method based on network tomography, and the network topology based on the statistical properties of the channel state estimation parameters. Traditional network tomography methods are based on the entire measurement cycle network status remains stable under the premise conditions topology estimation, but in the real network environment, the traffic on the link is bursty in nature,so the premise condition is not satisfied, so that the resulting estimates topology errors.Tomography method currently exists mostly use a metric to estimate the network topology, and individual metrics information contained in the rich enough, which makes use of only a measure of the network topology topology estimate was not accurate enough. To solve the above problem, this paper presents a topology can be estimated multi-topology estimation method in the non-stationary measurement network environment, the main work includes the following two aspects:1) Based on the first sequence to measure preferred search method: Depending on the destination node delay variation on a shared path is the same paper as an input signal by the delay using wavelet packet decomposition extract the metrics obtained multiple shared path length parameters, however, not all of the estimated measure of the network topology is useful, so we propose a measure based on the first sequence to the preferred method of searching through the use of pre-sequence to the search algorithm,using wavelet packet decomposition to obtain all get the metrics preferred to finally obtain an optimal set of metrics as the basis for multi-component metrics.2) Non-stationary network topology estimation method based on fuzzy clustering half: network tomography method currently exists mostly through node topology estimation fusion or hierarchical clustering method, and these methods have excessive or insufficient integration integration node the disadvantage, in order to accurately estimate the network topology, this paper presents a semi-topological estimation method based on fuzzy clustering, the method to measure more parameters as input to thedestination node dynamic clustering, and choose the best as a result of the clustering topology estimated network.In this paper, the proposed method of NS2 simulation. Simulation results verify the non-stationary network environment, characterized by a multi-metric choose to build in a non-stationary network environment to more accurately reflect the network of non-stationary time-varying characteristics. Meanwhile, the simulation results show that the network topology estimation method to measure more parameters based on the input half fuzzy clustering can effectively estimate the network topology, and get estimates than the network topology estimation method based on a more hierarchical clustering high estimation accuracy.This paper also use Planet Lab in the actual network environment to verify the correctness and validity of the method...