Using scalable double filtration algorithm to record more flows

摘要

For the limit of false positives probability in Bloom Filter (BF), this paper, according to the characteristic of network flows, proposed a novel mechanism based on scalable double filtration algorithm to record traffic information. The algorithm improves the performance from three aspects compared with the standard structure of Bloom Filter: using the timeout characteristics of TBF samples the mice flows, using CBF filtrate the packets of large flows according to the heavy-tailed distribution characteristics of network traffic, using dynamic multilayer scalable mechanism to reduce the false positives probability. The mechanism could not only record more flows, but also adapt dynamically when the number of flows rapidly rising. The experimental simulation results showed that compared with standard Bloom Filter, our algorithm can significantly improve the accuracy of recording flows information.