Research On Heavy Hitter Detection Based On Iterative Method In Software Defined Networking

Today traditional networking is more and more complicated,and it is difficult to configure and custom,so software defined networking is proposed and is attracting more and more interest from both researchers and network operators.Software defined networks have been deployed on a certain scale,and it plays a more and more important role today.The virtual network supported by software defined networking has enormously eased the deployment of the elastic computing and cloud services,and the technique to program the network has freed more and more operators from complicated and expensive component for network management.The services of software defined networking can be categorized as two classes: the network management services and the network measurement services.The network measurement is the foundation of the software defined networking,since only if we have the state of the network we can control it efficiently.The detection of Heavy Hitter or Elephant Flow is an important branch of the network measurement,through detecting and controlling Heavy Hitters most of the requirement of the traffic engineering,such as load balancing,QoS,congestion control and so on,can be satisfied.Heavy Hitters constitute most of the traffic volume in the network,if we can control the Heavy Hitters,we can control most of the volume in the network,but as the number of Heavy Hitters is small,the overhead to detect and control them is also small.This is the advantage of the detection of Heavy Hitter.Current studies about Heavy Hitter detection still focus on how to complete the detection in single switch.They do not consider how to do the detection using switches collaboratively in the real network.In this paper,we propose a distributed way to detect the Heavy Hitters,which is based on the iterative measurement.We also study how to improve the detection speed and assure the accuracy of the detection.Our main works are as follows:We propose a Heavy Hitter detection method,which can accept the tasks specified by users at real time.The system will dynamically allocate resource to these tasks to complete them.We study how to improve the detection speed and formulate the problem.We propose an algorithm to solve the problem and prove its correctness.We study how to assure the accuracy of the detection of Heavy Hitters.We implement a prototype of the measurement system.Through various experiments we prove that the system can improve the detection speed in a way and can assure the accuracy of the detection.