A survey of elephant flow detection in SDN
Binfeng Wang, Jinshu Su · 2018
Elephant flows usually transfer significant amount of data in the network, which may tend to cause network congestion and impact the network performance. Therefore, the accurate, timely and low-cost elephant flow detection is crucial in efficient network management. Taking advantages of SDN (software defined network), the elephant flow detection appears a fresh developing change. This paper firstly describes the related management mechanisms in SDN. Then according to the detecting location, the elephant detection approaches are classified into two kinds: detection in switch and detection in endhost. In switches, the detection approaches are well studied and can be not only based on the Flow-Statistics threshold but also based on flow characteristics. In end-hosts, the detection approaches are mainly based on the Flow-Statistics threshold and often collecting flow-statistics through monitoring socket buffer or hypervisor. Finally, the paper makes a performance comparison and puts forward several points of future prediction in terms of the elephant flow detection taking the deepening development of SDN into account.