Inferring and Querying the Past State of a Software-Defined Data Center Network

Abstract

Software-Defined Networking (SDN) is used widely in Data Center Networks (DCNs) to facilitate the automated configuration of network devices required to provide cloud services and a multi-tenant environment. The resulting rate of change presents a challenge to a DCN operator who needs to be able to answer questions about the past state of the network. We describe our work in addressing this need, and how an ontological approach was taken to build a topological and temporal model of a DCN, which could then be populated using control-plane data captured in a message log. Sophisticated queries applied against the populated model allow the DCN operator to gain insight into the effects of historical automated configuration changes. We have tested our model for accuracy against a network from which a message log was captured, and we have demonstrated how queries have been formulated to retrieve useful information for the DCN operator.