A compressive sensing-based reconstruction approach to network traffic

• We propose an approach to select the measured origin–destination flows.
• We use compressive sensing to reconstruct network traffic by part OD flows.
• Our method can decrease the estimation errors significantly.
• All OD flows can be recovered only by measuring 49–59% ones.

Traffic matrix in a network describes the end-to-end network traffic which embodies the network-level status of communication networks from origin to destination nodes. It is an important input parameter of network traffic engineering and is very crucial for network operators. However, it is significantly difficult to obtain the accurate end-to-end network traffic. And thus obtaining traffic matrix precisely is a challenge for operators and researchers. This paper studies the reconstruction method of the end-to-end network traffic based on compressing sensing. A detailed method is proposed to select a set of origin–destination flows to measure at first. Then a reconstruction model is built via these measured origin–destination flows. And a purely data-driven reconstruction algorithm is presented. Finally, we use traffic data from the real backbone network to verify our approach proposed in this paper.

Figure optionsDownload as PowerPoint slide