Spatio-Temporal Kronecker Compressive Sensing for Traffic Matrix Recovery

A traffic matrix is generally used by several network management tasks in a data center network, such as traffic engineering and anomaly detection. It gives a flow-level view of the network traffic volume. Despite the explicit importance of the traffic matrix, it is significantly difficult to implement a large-scale measurement to build an absolute traffic matrix. Generally, the traffic matrix obtained by the operators is imperfect, i.e., some traffic data may be lost. Hence, we focus on the problems of recovering these missing traffic data in this paper. To recover these missing traffic data, we propose the spatio-temporal Kronecker compressive sensing method, which draws on Kronecker compressive sensing. In our method, we account for the spatial and temporal properties of the traffic matrix to construct a sparsifying basis that can sparsely represent the traffic matrix. Simultaneously, we consider the low-rank property of the traffic matrix and propose a novel recovery model. We finally assess the estimation error of the proposed method by recovering real traffic.