An adaptive traffic measurement method for high-speed networks

High-speed network traffic is characterized by high burstiness and high randomness. Extensive test and analytical results show that high-speed network traffic has the statistical characteristic of Long-Range Dependence (LRD) or self-similarity. All traffic sampling measurement methods adopted currently are based on sampling algorithms in pure mathematical theories without consideration of the behavioral characteristics of actual network traffic, and affect the accuracy of network performance analysis. We present a sampling method of FARIMA-based traffic prediction, by which the sampling rate can be set dynamically based on the predicated traffic. The experimental results show that the sample can reflect the behavioral characteristics of traffic data more realistically.