Measuring the Control-Data Plane Consistency in Software Defined Networking

Abstract

Software Defined Networking (SDN) simplifies network management by separating the control plane from the data plane in performance networks. However, the actual packet behaviors, conforming to the rules in the data plane flow tables, may violate the original policies in the controller due to the inconsistency between the data plane and control plane. To address this problem, we propose 2MVeri, a new framework for measuring the control-data plane consistency, defined as the consistency between the control plane policies and data plane rules. 2MVeri uses a Bloom filter and a two-dimensional vector as a tag which is inserted in the packet header and is updated in each switch that the packet traverses. By exploiting path information compressed in the tag, 2MVeri can verify the control-data plane consistency. Moreover, when verification fails, 2MVeri can localize the faulty switch. Evaluations conducted on a datacenter network with the fat tree topology k=4 demonstrate that 2MVeri achieves 100% accuracy in consistency verification and high fault localization performance.