A Study of the Forwarding Blackhole phenomenon during Software-Defined Network Updates

Software-Defined Networks (SDNs) provide a programmable and flexible framework to frequently reconfigure networks, e.g. updating forwarding policies. Nonetheless, such updates should be performed consistently and efficiently to support high-scale network topologies. Whereas SDN provides a centralized and direct control entity (the controller) to manage networks, SDN remains a distributed system. Indeed, to perform an update, the controller and switches should communicate via message-exchange through an asynchronous communication interface. This inevitably brings a variance with respect to message transmission time as well as rule installation time on the switches. With the interleaving of in-fly packets and update operations, consistency can be harmed, leading to connectivity inconsistency updates, i.e. problems related to packet delivery disruption. Forwarding Blackhole (FB) is one of the problems leading to such update inconsistencies. Once occurred, packet delivery can be disrupted and even worse, packets can be lost. In this paper, we present an analytical study of the FB connectivity inconsistency phenomenon during SDN updates. We identify the three cases in which FBs can occur, and we present them as FB patterns. Based on this study, we propose a novel update SDNs approach, in which update operations are evaluated on the switches themselves, instead of being handled in the controller as traditional approaches. Finally, we discuss how the proposed approach may enhance the update performance.