A Framework for Classification and Visualization of Elephant Flows in SDN-Based Networks

Abstract

Long-lived flows termed as elephant flows normally transport large volumes of data in enterprise networks, particularly data center networks. These flows tend to consume a lot of bandwidth and fill up network buffers end-to-end. This causes non-trivial delays for short-lived flows referred to as mice flows which are usually delay-sensitive. Therefore, identifying and handling elephant flows is important for QoS provisioning. In this paper, we present a framework for real-time detection and visualization of elephant flows in SDN-based networks using sFlow. Using our proposed framework, network operators can examine elephant flows through each switch by double-clicking the switch node in the topology visualization UI. Although not in the scope of this paper, but in order to meet traffic engineering requirements, the elephant flows detected and visualized by our proposed framework can be reprioritized, re-scheduled, or routed via dedicated high speed links. We evaluate the proposed framework by using a physical SDN testbed as well as a Mininet-based testbed.