The Comparison of Split-Flow Algorithms in Network Traffic Classification: Sequential Mode vs. Parallel Model

Abstract

In recent years, with the popularity of the network, the assistance of the network seem to be much more important. It requests much higher performance to the network as well as great pressure to available bandwidth. And P2P (Peer-to-Peer) network accelerates the exhaustion of the rest bandwidth. Therefore, different network traffic classification algorithms turn up one after another, such as machine learning, data mining, pattern recognition etc. No matter which algorithms are adopted, one technique detail cannot be neglected, classifying the packets into flows (flow-split) process. This process may achieve better time complexity and space complexity in short time interval traffic traces. But with the extension of the traffic traces scale, the split-flow process appears time-consuming and effort-consuming. In addition, split-flow plays irreplaceable role in the construction of the classification algorithms and the classification. In this paper, we propose parallel flow-split model and estimate the efficiency with the sequential model to derive our experimental results.