EXPRL:experience and prediction based load balancing strategy for multi-controller software defined networks
| Autor: | D. M. Akbar Hussain, Anuradha Banerjee |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2022 |
| Předmět: |
Overloaded controller
Computer Networks and Communications Computer science Industry standard Throughput 02 engineering and technology Artificial Intelligence Packet loss 0202 electrical engineering electronic engineering information engineering Forwarding plane Electrical and Electronic Engineering Latency (engineering) Predicted call arrival rate Target switch business.industry Applied Mathematics Target controller 020206 networking & telecommunications Load balancing (computing) Computer Science Applications Network planning and design Computational Theory and Mathematics Latency Actual call arrival rate 020201 artificial intelligence & image processing business Software-defined networking Load balance Information Systems Computer network |
| Zdroj: | Banerjee, A & Hussain, D M A 2022, ' EXPRL : experience and prediction based load balancing strategy for multi-controller software defined networks ', International Journal of Information Technology (Singapore), vol. 14, no. 4, pp. 2155-2169 . https://doi.org/10.1007/s41870-019-00408-5 |
| DOI: | 10.1007/s41870-019-00408-5 |
| Popis: | Software Defined Networks or SDN has proved itself to be backbone in a network design of the new era and is quickly becoming an industry standard. It allows decoupling of control and data plane for efficient monitoring of the network traffic. These networks can be divided into two main classes—single and multiple controller SDN. A single controller is unable to control flows if the number of switches connected to it increases up to a great extent and traffic generated by them overload the controller. In that case a pool of multiple controllers is a necessity. There each controller is initially connected to approximately same number of switches. Still, if a controller feels overloaded then some of its switches are disconnected from it and connected to some comparatively under loaded controller to balance load in the network. The present article EXPRL proposes an experience and prediction based load balancing strategy that efficiently identifies overloaded controllers and selects target underloaded controllers to shift some load from the overloaded ones. Simulation results show that EXPRL enables the network to greatly increase network throughput and reduce network latency as well as migration cost than its state-of-the-art competitors. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|