Routing Multiple Flows and Collecting Link State Information in SDN
Software-Defined Networking (SDN) is a new paradigm that attempts to address various challenges in traditional networks by separating the control plane and data plane. Potential benefits of SDN have been widely studied in data centers and local area networks (LANs). Naturally, researchers are now asking how to take advantage of this innovative technology in the context of wide area networks (WANs). However, because of physical differences between LANs and WANs (e.g., large number of geographically dispersed nodes, long propagation delays), several new challenges arise such as where to place the SDN controllers, how to synchronize their actions, how to accurately and quickly obtain the global view and state information of the network, how to use this information in making routing decisions to support various applications with QoS requests.
In this dissertation, we mainly focuses on the last two challenges and propose new mechanisms in the context of SDN-based WANs. Assuming the availability of accurate state information (as commonly done in the literature), we first investigate various path selection algorithms that can deal with multiple unsplitable flows with QoS guarantees while also providing load-balancing. Since the underlying path selection problems are NP-hard, we focus on heuristic methods and demonstrate their efficiency through simulation-based studies. We then question the assumption of having accurate state information and investigate how the routing performance would be impacted in practice, where the state information is collected periodically to minimize protocol overhead. Based on our findings, we develop a new link-prioritized network state collection mechanism and demonstrate its efficiency through Mininet emulator. The proposed solutions advance the field by enabling the development of new routing mechanisms to deal with congestion, load balancing, and traffic management in the Internet.