Ensuring data plane consistency in SDN-based publish/subscribe systems

Abstract

Content-Based publish/subscribe paradigm is a widely used paradigm which enables applications to share events, and allow the integrated applications to remain loosely coupled. The importance of the paradigm increases further with the emergence of IoT applications, a micro service approach to application development, etc. With the rise in the amount of data being shared between applications, there is a need to provide high data rate and bandwidth efficiency, and the introduction of Software Defined Networking(SDN)-based pub/sub, it becomes possible to achieve both. Content-based pub/sub built on SDN uses the IP address match fields of the flows in Ternary Content-Addressable Memory(TCAM) as content filters. While such an in-network filtering offers high data rate, it also involves frequent updates in the data-plane to keep the system running optimally. Although pub/sub using SDN is efficient, to use it in real-time, it needs to ensure some level of consistency when updating content-filters in the data plane so that events are not subjected to black holes, duplicates, loops, etc. Keeping the requirement of ensuring consistency in mind, the goals of this thesis include an analysis of the data plane consistency issues related to SDN-based pub/sub, to apply existing algorithms or implement new algorithms to preserve consistency of the data plane, and to provide an evaluation comparing all the implemented solutions.