Research On High-speed Network Flow Measurement Model

The measurement model based on flows is the important direction and basement of research on high-speed network flow. The improvement and optimization of algorithms are very important for validition of flow measurements. This thesis introduced the last research results in international flow measurement algorithms. Based on these researches, it illuminated the difficulties and research directions of flow measurement algorithms. This result could influence the more deep research in flow measurement algorithms in positive.Based on Bloom Filter algorithm, this thesis introduced an algorithm Multi-Dimensional Bloom Filter(MDBF), which could be used to calculate the frequencies of different items in a set using one of its characteristic is long-tailed distribution on that set. Because the distribution of flows length in network is long-tailed, the MDBF algorithm was very fit for flow informations'statistics and analysis. For the characteristic of network measurement, this thesis presented the methods of fixed time renew and repeated minimum value to optimize the algorithm, which could reduce the calculation errors. And then this thesis designed the high-speed network flow measurement model based on MDBF. The prototype of this model was realized by C language, which expatiated usage of measurement model.This thesis verified the efficiencies of measurement model using the datasets coming from the PMA group of NLANR. The experiment results indicated that this measurement model could be used to not only find out the long flows which occupy most resource of network, but also recognize the worm attacks and DDoS attacks in the network efficiently. The experiment results proved the flow measurement model designed by this thesis can be used in practicality.