Research On Fine-Grained Measurement Technique In Software-Defined Networking

Network measurement comprise several different aspects including traffic measurement,performance measurement,failure measurement and so on,among which traffic measurement stands as the most important one.Traffic measurement approaches count the statistics of traffic transmitted in network.It can provide essential inputs for various network management tasks.With the development of network technologies and applications,the size and complexity of network is increasing and the number as well as the size of traffic also increase accordingly,which makes it is surprisingly difficult to measure the network traffic accurately.What makes the situation even worse is that,in traditional networks,switches can only process and forward packets according to fixed protocol rules,which leads to the lack of flexible control and measurement ability.Fortunately,empowered with a flexible and global control plane as well as a programmable data plane,Software-Defined Networking(SDN),proposed in recent years,comes as a salvation which lays a foundation for fine-grained traffic measurement.Therefore,this paper mainly focuses on fine-grained network traffic measurement in SDN.Single-point traffic measurement refers to the method of measure the traffic data passing through the node on a single network node.This is the first main aspect this paper tries to shed some light on.Current single-point flow measurement methodology tends to mix elephant and mice flows and thus results in low throughput and high measurement error.So this paper presents a high performance single-point flow measurement algorithm with high performance by carrying out measurement for the elephant and mice flow respectively,with the high-speed TCAM for accurate measurement of elephant flow and sketch for rough measurement of mice flow.This paper argues that this kind of separation brings about a huge improvement of accuracy and throughput.This is also demonstrated in experiments.Experiment results show that method proposed in this paper can detect and estimate the flow accurately,yielding a 430% thoroughput improvement compared with traditional approaches.Network-wide traffic measurement refers to the method of multi-node coordinated measurement of the entire network traffic within the whole network.Regarding this,aiming at making full use of the measurement resources and improve the accuracy as well as throughput of the measurement,this paper mainly focuses on two key problems,the deployment of the measuring nodes and the assignment of the measurement tasks.First of all,this paper proves the deployment of measuring node is a NP-hard problem and then give an approximate solution.To avoid repeated measurement and perform measuring task assignment given the constrained measurement node capacity,this paper proposes a task assignment algorithm based on the minimum cost.The experiment shows that the measurement task assignment method can reasonably utilize the network measurement resources and effectively reduce the measurement error.In this paper,the elephant flows are measured by TCAM entries,which can also be used to adjust the routes of flows.Based on this observation,this paper proposes a joint optimization method for traffic measurement and routing.The joint optimization method adjusts the traffic routing strategy based on the sizes of elephant flows measured by TCAM entries.To get the optimal routing strategy,this paper first formulates the proplem as a MILP(Mixed Integer Linear Programming)model with the optimization objective of load balance,and then proposes a heuristic algorithm to solve the problem efficiently.Experiments show that the proposed heuristic algorithm can get near-optimal routing solution in a short time.