Research On Performance Data Compression Processing For Heterogeneous Network

In order to guarantee the heterogeneous network is always in a state of run efficiently, networkmanagement system requires real-time monitoring the entire operation information of theheterogeneous network. The complex structre and various equipments will cause massiveperformance data, which on the one hand, increase the pressure of data transmission in networkmanagement system, even reduce the transmission efficiency of performance data; on the otherhand will increase the cost of server storage data. Though performance data collected in thecontinuous time not only itself has time redundancy, but performance data of different attribute alsohave certain relevance, considering the time redundancy and multivariate redundancy, this paperputs forward appropriate data compression methods, to remove the performance data redundancyand decrease the amount of performance data transmission.Firstly, considering the time redundancy of performance data itself, the compressed sensingalgorithm based on wavelet transform is adopted. Performance data can be divided into the highfrequency and low frequency by wavelet transform, Because the high frequency meet therequirement of compressed sensing sparse, so it’s dealled by compressed sensing, namely withfewer measurements to show more high frequency coefficient. When reconstruct the high frequencycoefficient on the server side, this paper proposes an algorithm of SAMP reconstruction based onthe wavelet tree. the experiments verify the algorithm not only have high compression rate when inthe premise of accurate reconstruction, but also greatly improve the accuracy of the reconstructionalgorithm, reduce the number of iterations of reconstruction, decrease the running time of thereconstruction.Secondly, considering from the angle of Multivariate redundancy between performance data,low frequency component is approximation approach of performance data, the correlation betweendifferent performance data still exist in the low frequency component, so this paper proposes aMultivariate compression based on segmented SBR algorithm to further compress the lowfrequency component, remove the correlations between different attributes. Through experiments itis concluded that the method on the basis of meet the error limit, greatly reducing the amount of lowfrequency component of the transmission, further reducing the amount of performance datatransmission.Finally combine with the compressed sensing based on wavelet transform and segmented SBR algorithm, considering the compression performance of the two algorithms on overall performancedata. In the experiments, the two kinds of algorithms are not only good to remove the redundancyperformance data, reduce the throughput performance data, but also have high reconstructionaccuracy.