Survey on Reinforcement Learning based Efficient Routing in SDN

Quach, Hong-Nam; Yeom, Sungwoong; Kim, Kyungbaek

doi:10.1145/3426020.3426072

Cited by 2 publications

(1 citation statement)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…computer network consists of many pieces of equipment including routers, switches, and firewalls, each functioning according to highly complicated protocols. In traditional networks, each one of the devices that act on different network layers may have been purchased from a different company, and the variety of their rules and structures can complicate network management actions [1]. For this reason, it is difficult to implement management policies in traditional networks and novel management ideas cannot be easily applied in a consistent and homogeneous manner, thereby necessitating the development of new types of networks to overcome these problems.…”

Section: Introductionmentioning

confidence: 99%

Traffic-aware Routing with Software-defined Networks Using Reinforcement Learning and Fuzzy Logic

Jaafari¹,

Nassiri²,

Mohammadi³

2022

IJC

View full text Add to dashboard Cite

In recent years, the idea of software-defined networks (SDNs) has been proposed for better network management. This architecture has succeeded in optimizing network management functions and increased the ability to synchronize network equipment. Currently, one of the major issues in this architecture is the routing of packets flowing in the network. The main aim in the routing of packets is to increase the quality of services. Enhancement of the quality and productivity of these networks will increase user satisfaction. To this end, the present study proposes a mechanism for selecting the best route from among several existing routes to direct a flow on such a network. The proposed method examines the network parameters including bandwidth, delay, and packet loss on each link of the route by using artificial intelligence algorithms and changes the parameters reducing network productivity by means of fuzzy logic. Our evaluations show that the proposed method can select routes with high productivity and increase the quality of services on the network. Receiving feedback and modifying the fuzzy membership functions related to each mentioned criterion can maintain the effect of these parameters on an acceptable level after which all transmissions tend towards the optimum. Given the use of reinforcement learning methods which underpin some of the routing methods in SDNs, the proposed idea may gradually contribute to the provision of optimized services on the network.

show abstract

Section: Introductionmentioning

confidence: 99%