A Dynamic Load Balancing Algorithm for Data Plane Traffic
Lucas Soares da Silva, Carlos Renato Storck, Fátima Duarte-Figueiredo · Zenodo (CERN European Organization for Nuclear Research) · 2019
There is a worldwide growing demand for high bandwidth, low latency and reliability. Software Defined Network (SDN) emerged a few years ago as a new paradigm, separating the control and the data planes to improve the network management. The SDN concept applies to any type of network and is indispensable in new generations of networks such as 5G. One of the main problems in the data plane is the distribution of traffic. Quality of service (QoS) characteristics such as throughput and latency of each application can be affected by inefficient link balancing. This paper proposes an algorithm for dynamic load balancing the data plane traffic. The load balancing minimizes the effects of intense network data traffic, mitigating bottlenecks. The proposed algorithm dynamically changes the links flows as the network utilization intensifies. The algorithm finds the shortest paths and calculates the link cost choosing the best link after identifying and kind of bottlenecks. This reduces the latency and the packet loss in scenarios with network congestion. Experiments were conducted using the Mininet and the OpenDayLight controller. Through emulation results, it wasobserved that there was a significant improvement in jitter, packet loss, and throughput. Thus, the proposed algorithm can work in any kind of SDN environment, balancing the data plane traffic flow.