Research On Traffic Measurement And Routing Optimization For Programmable Network

Network traffic measurement can provide necessary data input for network management tasks such as anomaly detection,fault analysis,congestion control,routing optimization,etc..The traditional network devices can only deal with the data packets according to the fixed protocol rules,which leads to the lack of flexible control and measurement capability.The exchange equipment in the programmable network has the programmable characteristic,can carry on the processing to the data packet flexibly through the programming,has established the foundation for the implementation fine granularity network flow measurement.Therefore,this thesis mainly studies the flow measurement and management method for programmable network.Because the switches in a programmable network are programmable,fine-grained measurements of the flow can be made independently at each node.Due to the limited computing and storage resources of nodes,it is not enough to measure all flows accurately.At present,single-point flow measurement methods measure the size flow separately,use counter to measure the large flow accurately,and use Sketch with data compression ability to measure the small flow roughly.However,these methods still have some deficiencies in the accuracy of extracting the main flow,if the main flow can be extracted more accurately,the measurement accuracy can be improved effectively.Therefore,a single-point traffic measurement method based on counter and Sketch,Flexible Sketch,is proposed in this thesis.Experiments show that the Flexible Sketch has lower measurement error under the same memory capacity.A flow in a network usually passes through more than one node,and if measurements are made on each node,repeated measurements can result.Network-level traffic measurement is implemented through the cooperation of multiple nodes,each node measures a different flow,which avoids repeated measurement and improves the utilization of measurement resources.Furthermore,network-level traffic measurement can reduce the number of measured traffic on a single node and improve the measurement accuracy.In order to achieve network-level traffic measurement,the key measurement node assignment problem is studied in this thesis.Firstly,an approximate algorithm based on Joseph-Louis Lagrange relaxation is proposed to solve the problem.However,the computational complexity of the algorithm is positively related to the size of the network.In view of this,a genetic algorithm is proposed to solve the problem.The experimental results show that the two methods can solve the optimal solution effectively.With the rapid growth of internet traffic,routing optimization has become a very important aspect of network management.Through route optimization,the load balance of network traffic can be realized and the network performance can be improved.However,the current routing optimization method is mainly based on local link information,missing the consideration of the whole network traffic information,the actual effect is not ideal.In view of this,this thesis proposes a routing optimization method based on network traffic measurement,which obtains the whole network traffic information through network traffic measurement,and extracts the main flow in the network for individual routing optimization,for better optimization.Experimental results show that the proposed algorithm can effectively reduce the maximum link utilization and achieve load balancing.