Multi-level Interval Based Load Balancing for Multi-controller SDN
No Thumbnail Available
Date
2023-10
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Software Defined Networking (SDN) is a new approach of managing and programming networks that is managed by a centralized controller. For a larger network, single controller architecture will be inefficient to manage the network because the increase in the network traffic causes network congestion and transmission delays. To tackle this issue, software defined architecture with multiple controllers has been introduced in recent researches. However, this strategy introduces a new problem of unequal load balance on the distributed SDN controller. To address this issue, several load balancing approaches have been developed. But most of them are domain specific and have relatively high migration cost, communication overhead and packet loss with low load balancing rate. An Improved Switch Migration Decision Algorithm (ISMDA) is one of the algorithm that solves the SDN load imbalance when the incoming data load is high traffic (traffic load greater than 500 p/s). The under loaded controller is selected by the load balancing module among the controller group by estimating variance and the average load status of the controllers. It improves controller throughput, migration cost and response time by running different modules like load judgment, switch selection and controller selection module. But it splits the load only into two intervals, i.e., overloaded and under-loaded, which limits its capability to distinguish the controller load status accurately. This paper proposes Multi-level Interval Based Load Balancing (MIBLB) algorithm that modifies the ISMDA algorithm by implementing a multi-level load interval, i.e., idle, normal, overloaded and highly overloaded, and applies dynamic threshold for switch migration. Dividing the load status into multi-level intervals enables accurate migration by giving priority to highly overloaded controller, and avoid congestion and packet loss. The efficiency will be evaluated in terms of network performance evaluation parameters, i.e., throughput, response time, and average number of migrations. The experimental results show that MIBLB reduces number of migrations and controller response time of ISDMA by 6.43% and 3.53%, respectively. The controller throughput of ISDMA is also increased by 6.82% while sustaining efficient load balancing rate.
Description
Keywords
Load balancing algorithm, Software defined networking, multi-controller, switch migration.