ReFeR: Resource Critical Flow Monitoring in Software-Defined Networks

• Published in: 2018 IEEE Global Communications Conference (GLOBECOM) pp. 1 - 7
• Main Authors: Yihui Qian, Yutong Liu, Linghe Kong, Min-You Wu, Mumtaz, Shahid
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.12.2018
• Subjects: Computer crime; Control systems; Error analysis; IP networks; Monitoring; Particle separators; Task analysis
• ISSN: 2576-6813
• DOI: 10.1109/GLOCOM.2018.8647445

Flow monitoring is widely applied in softwaredefined networks for monitoring network performance. Especially, the detection on heavy hitters can prevent the Distributed Denial of Service attack. However, many existing approaches fall in one of two undesirable extremes: (i) inefficient collection where only accuracy is concerned in the method; (ii) low accuracy caused by the sacrifice with fast detection. As a result, we aim to find a balance between the accuracy and efficiency of flow monitoring, where the network resources can be saved and the error rate can also be confined simultaneously. In this paper, we present ReFeR, a novel "Report-FeedbackReport" scheme to improve the detection efficiency of heavy item detecting while ensuring low error rate of the measurement. ReFeR leverages the binary order of magnitude of item measurement to replace the long statistical information shared between switches and controller; after the items are analyzed with the magnitude, only fewer uncertain items are involved in further detection, where their information (i.e., significant digits) is provided for final judgment. Theoretical analysis and simulated evaluation have proved the effectiveness of our solution. ReFeR keeps the error rate under 1% and the saving rate larger than 20% in most cases as the selecting fraction α> 524, which guarantees both high efficiency and low error rate compared with existing methods.