Electronic Thesis and Dissertation Repository

Thesis Format

Integrated Article


Master of Science


Computer Science


Anwar Haque


Traffic routing is vital for the proper functioning of the Internet. As users and network traffic increase, researchers try to develop adaptive and intelligent routing algorithms that can fulfill various QoS requirements. Reinforcement Learning (RL) based routing algorithms have shown better performance than traditional approaches. We developed a QoS-aware, reusable RL routing algorithm, RLSR-Routing over SDN. During the learning process, our algorithm ensures loop-free path exploration. While finding the path for one traffic demand (a source destination pair with certain amount of traffic), RLSR-Routing learns the overall network QoS status, which can be used to speed up algorithm convergence when finding the path for other traffic demands. By adapting Segment Routing, our algorithm can achieve flow-based, source packet routing, and reduce communications required between SDN controller and network plane. Our algorithm shows better performance in terms of load balancing than the traditional approaches. It also has faster convergence than the non-reusable RL approach when finding paths for multiple traffic demands.

Summary for Lay Audience

In the past decades, the number of the Internet users and the type of services that rely on the Internet have increased dramatically. Traffic routing, the process of sending data from the source to its specified destination, is vital for the proper function of the Internet. By combing novel network architecture and techniques, researchers hope to develop routing algorithms that provide better performance than traditional approaches.

Based on Reinforcement Learning (RL) and Segment Routing (SR), we developed a routing protocol, RLSR-Routing, over Software Defined Networking (SDN). RLSR-Routing can self-explore the network and find a path for a given traffic demand based on user-defined optimality. Compared with previous work, our approach adopted some modifications to RL algorithm, such that the cost of finding the path is minimized. In addition, our approach can reuse previously learned knowledge about network status, when it is working on new traffic demands. In experiment settings, RLSR-Routing outperforms non-RL based routing algorithm in terms of load-balancing among network links. Compared with the RL approach that does not reuse previous learning results, RLSR-Routing shows faster convergence speed.