Twitter Feed
Couldn't connect with Twitter

What are you looking for?

Simply enter your keyword and we will help you find what you need.

DQR: Deep Q-Routing in Software Defined Networks

Software-defined networking is gaining popularity day by day. There exist different types of SDN controllers that enable network administrators to dynamically optimize network resources and provide flow-level quality of service (QoS) provisioning. This QoS routing in SDN needs to be investigated to improve end-to-end throughput. Traditional methods exist but those methods have certain limitations when dealing with multiple QoS parameters.

To deal with multiple QoS parameters and to solve this complex problem DRL methods have become very popular to be used. Because these methods have the ability to learn from experience that allows avoiding the development of large accurate mathematical models. But due to the use of Q-learning, different DRL approaches do not perform well when the routing complexity increases.

To solve the problem, this project aims to investigate the task of quality of service (QoS) routingin software-defined networks (SDN).

Proposed Methodology
To achieve the above-mentioned goal, this project uses different QoS parameters such as delay,bandwidth, loss, and cost. For this purpose, a new DRL solution for greedy online QoS routing in SDN is implemented. This DQR utilizes a dueling deep Q-network with prioritized experience replay to compute a path for any source-destination pair request in the presence of multiple QoS metrics. In this case, we consider SDN which consists of three layers. This architecture of SDN is mentioned in the Figure below.

Results and Conclusions
In order to validate the simulation, the experiment benchmark DQR against two other greedy online routing methods is performed. For this purpose, two different communication topologies were taken under consideration. From the results, it is evident that DQR can be used for greedy online QoS routing and that it achieves better performance than QAR and SP with respect to different QoS metrics by efficiently utilizing network resources. Moreover, it is also concluded that the design of DQR is flexible enough that it can be used for any network topology and it can be used for a different number of QoS metrics.




Previous Project
Next Project