A Low-Cost and Resilient Flow Monitoring Framework in SDN

Abstract

Network measurement and monitoring are essential for learning the current network state and act accordingly. Moreover, Software-Defined Networking (SDN) makes the measurement and monitoring more accessible and flexible. However, existing measurement schemes in SDN suffer from high measurement cost due to a fix sampling rate while monitoring all the data plane elements. In addition, existing monitoring schemes are not resilient in the presence of communication link and node failures. Therefore, in this thesis, we propose a low-cost and resilient flow monitoring framework in SDN. We first propose a low-cost measurement algorithm, which reduces the measurement cost by aggregating flows at a subset of switches. Next, we define a model to optimization the measurement cost and accuracy. Furthermore, we observe that link and node failures can impact measurement accuracy. Thus, we propose a resilient monitoring framework called ReMon. In particular, we propose three algorithms to recover from node and link failures, which are implemented in the SDN controller. Then, we update the measurement scheme after recovering from a failure. The experimental results show that the proposed solutions outperform their counterparts in terms of measurement and computation cost, accuracy, recovery time, and memory usage.