Data reconstruction in internet traffic matrix

摘要: Traffic matrix is an abstract representation of the traffic volume flowing between sets source and destination pairs. It a key input parameter network operations management, planning, provisioning engineering. also important in context OpenFlow-based networks. Because even good measurement systems can suffer from errors data collection fail, missing values are common. Existing completion methods do not consider exhibit characteristics only provide finite precision. To address this problem, paper proposes novel approach based on compressive sensing self-similarity to reconstruct flow data. Firstly, we analyze real-world matrix, which all low-rank structure, temporal smoothness feature spatial self-similarity. Then, propose Self-Similarity Temporal Compressive Sensing (SSTCS) algorithm The extensive experiments with show that our proposed SSTCS significantly reduce reconstruction achieve satisfactory accuracy comparing existing solutions. Typically successfully less than 32% when as much 98% missing.