Detection of Super Nodes for Network Data Streams with Filtering Strategy

Due to the massive number of network data streams in high speed networks, it is not feasible to maintain the exact connection degree of each node in the limited memory and computation resource. According to the low accuracy and memory utilization for detection of super nodes based on flow sampling, we propose detection of super nodes based on filtering strategy for multiple data streams (DSF). The main advantage of our approach is that it can eliminate the majority of nodes with potentially low connection degree and preserve nodes with large connection degree in the measurement interval, so it can reduce the memory wasted by nodes with low connection degree. Our approach is mainly consist of packet preprocessing, updating and merging, filtering strategy, super node detection. The experiments are conducted based on the real network traffic and the testing results show that the proposed method outperforms CSE (Compact Spread Estimator) and JDS (Joint Data Streaming and Sampling) in terms of detection accuracy, storage utilization and time overhead.