An OpenFlow-Based Load Balancing Strategy in SDN

Xiaojun Shi,Yangyang Li,Haiyong Xie,Tengfei Yang,Linchao Zhang,Panyu Liu,Heng Zhang,Zhiyao Liang
DOI: https://doi.org/10.32604/cmc.2020.06418
2020-01-01
Abstract:In today's datacenter network, the quantity growth and complexity increment of traffic is unprecedented, which brings not only the booming of network development, but also the problem of network performance degradation, such as more chance of network congestion and serious load imbalance. Due to the dynamically changing traffic patterns, the state-of the-art approaches that do this all require forklift changes to data center networking gear. The root of problem is lack of distinct strategies for elephant and mice flows. Under this condition, it is essential to enforce accurate elephant flow detection and come up with a novel load balancing solution to alleviate the network congestion and achieve high bandwidth utilization. This paper proposed an OpenFlow-based load balancing strategy for datacenter networks that accurately detect elephant flows and enforce distinct routing schemes with different flow types so as to achieve high usage of network capacity. The prototype implemented in Mininet testbed with POX controller and verify the feasibility of our load-balancing strategy when dealing with flow confliction and network degradation. The results show the proposed strategy can adequately generate flow rules and significantly enhance the performance of the bandwidth usage compared against other solutions from the literature in terms of load balancing.
What problem does this paper attempt to address?