Low cost network traffic measurement and fast recovery via redundant row subspace-based matrix completion

Traffic matrices (TMs) are essential for managing networks. Getting the whole TMs is difficult because of high measurement cost. Several recent studies propose sparse schemes to reduce cost, which involve taking measurements on only a subset origin and destination pairs (OD pairs) inferring data unmeasured OD through matrix completion. However, existing network suffer from problems computation costs low recovery quality. This paper investigates coherence feature real traffic flow traces (Abilene GÈANT). Both sets coherence, with column greater than row coherence. According both sets, we our Redundant Row Subspace-based Matrix Completion (RRS-MC). RRS-MC involves several techniques. Firstly, design an algorithm identify subspace rows historical data. Secondly, based identified rows, sampling scheduling algorithm, takes full samples in while partial remaining rows. Moreover, redundant rule prevent accuracy decrease caused by varying. Finally, completion recover partially measured We conduct extensive experiments. Results indicate that proposed scheme superior state-of-the-art accuracy.