Detecting heavy flows in the SDN match and action model

摘要: Abstract Efficient algorithms and techniques to detect identify large flows in a high throughput traffic stream the SDN match-and-action model are presented. This is contrast previous work that either deviated from match action by requiring additional switch level capabilities or did not exploit data plane. Our construction has two parts; (a) new methods efficiently sample an model, (b) efficient scalable way, model. flow detection provide accuracy present good practical tradeoff between - controller traffic, number of entries required table. Based on different parameters, we differentiate heavy flows, elephant bulky types. Additionally, as part our scheme, sampling packets with arbitrary probability p per packet byte traverses switch. Finally, show how can be adapted distributed monitoring setting multiple switches, easily scale switches.