DW-Sketch: A Sketch-based scheme for realizing multi-network measurement tasks

Abstract

Network measurement includes many aspects of tasks including heavy flows detection, attack and anomaly detection, traffic statistics and so on. Under high-speed network traffic, in order to process huge data efficiently in real time for realizing network measurement, many sketch-based techniques with building a small data-structure to store all of flows have been proposed in related research. Existing sketch-based techniques were implemented by proposing complex algorithms customized to specific measurement task. Actually, Network administrators need to learn the network state from different perspectives, so it is necessary to build a general sketch architecture that can accomplish multiple measurement tasks in parallel. Based on this, this paper proposes DW-Sketch which realizes four network measurement tasks, including heavy hitter detection, heavy changer detection, flow size estimation and flow cardinality estimation. Compared with existing sketches for different measurement tasks with experiments, DW-Sketch is proved to be a multi-task network measurement technique with good measurement effect through certain evaluation indexes, such as precision, relative error, etc.