SARD: A Smart Approach of Rule Division for fast flow-level consistent update in SDN

Abstract

In Software Defined Networks (SDN), the configuration inconsistency during updates is one main source of network instability. Even if the validity of the initial and final configurations is guaranteed, the interim complex and inconstant network states might cause routing conflicts and transmission disruption. Therefore, an efficient updating scheme with configuration consistency is required. Current schemes well guarantee the packet-level consistent update, but perform poorly for the flow-level case.In this paper, we present a Smart Approach of Rule Division (SARD) for fast flow-level consistent update in SDN. We first provide a simplified mathematical model of the network. Based on this model, we then propose SARD to guarantee the flow-level consistency during configuration updates. In SARD, the controller (1) collects the information of existing flows; (2) computes K optimal prefixes covering these flows, meanwhile with the minimized space; (3) installs the new rule and K old sub-rules with lower and higher priorities respectively. SARD preserves the flow-level consistent property and accelerates the process of configuration update in SDN. We evaluate the performance of SARD by comprehensive experiments. The results show that our scheme reduces the transition time to about 10% of the current method of periodical direct division.