Elephant flow detection and delay-aware flow rerouting in software-defined network

Abstract

Long-lived large (elephant) flows carry a large amount of data and consume a lot of network resources, such as VM migration and backup data transfer. Normal (mice) flows carry a small amount of data and consume a few network resources, such as web searching. The long existence of elephant flow makes the network throughput degradation and delays to other traffic flow in the network by utilizing more bandwidth and buffers. Therefore, differentiating elephant and mice flow plays as an important function to optimize network performance. This paper presents a traffic rerouting solution with the combination of elephant flow detection and multipath routing in software-defined networking (SDN). This solution provides network performance improvement and efficient resource utilization by rerouting the elephant flow to the least delay path among available paths. The proposed method is implemented by using ONOS controller and Mininet emulator. The experimental results show that 44.11%- 58.76% throughput improvement and 36.46%-47.44% flow completion time (FCT) reduction for elephant flows as compared with the reactive forwarding application of ONOS controller.