A New End-To-End Network Traffic Reconstruction Approach Based on Different Time Granularities

Abstract

AbstractEnd-to-end network traffic is an important input parameter for network planning and network monitoring, which plays an important role in network management and design. This paper proposes a new end-to-end network traffic reconstruction algorithm based on different time granularity. This algorithm reconstructs the end-to-end network traffic with fine time granularity by taking advantage of the characteristics of the link traffic which is easy to be measured directly in the network with coarse time granularity. According to the fractal and self-similar characteristics of network traffic found in existing studies, we first carry out fractal interpolation for link traffic measurement under coarse time granularity to obtain link traffic under fine time granularity. Then, by using the compressive sensing theory, an appropriate sparse transformation matrix and measurement matrix are constructed to reconstruct the end-to-end network traffic with fine time granularity. Simulation results show that the proposed algorithm is effective and feasible.KeywordsEnd-to-end network traffic reconstructionFractal interpolationCompression senseDictionary Learning Algorithm