基于数据到达间隔的网络大流检测方法OA

Detecting elephant flows is a critical task in network monitoring and management.The existing detection methods main-ly focus on the size of the flows,where they filter the elephant flows by size.However,the real elephant flows in high-speed net-work tend to be swamped by a large number of mouse flows,especially under limited memory size,resulting in the algorithm not being able to correctly detect elephant flows by size.In this paper,we propose a novel method which separates the elephant flows from the mouse flows by expelling the oldest flow which also takes on big arrival interval of data.Experimental results show that the new method achieves good precision with limited memory size.