Measurement And Analysis Of Network Traffic Performance Based On Stochastic Network Calculus

The rapid growth of wireless communication services has put forward higher requirements for network performance.Queuing theory is a traditional analysis method of network.With the complexity of network structure and the diversity of services,this theory has its limitations.As a new analysis method of network performance,network calculus can make up for the limitations of queuing theory in theoretical modeling and performance research.Therefore,it can give more accurate performance prediction in network queuing issues.At present,the bottleneck of network calculus application lies in the difference between theoretical model and real network,which leads to deviation of theoretical analysis.Therefore,model analysis needs to be tested for subsequent revision.However,the test of network calculus is usually based on simulation.The performance boundary analyzed by theory is often difficult to verify in the actual test because performance and precision of test equipment are often limited.It is difficult to find external influences of network through simulation only.In addition,for scenarios with strong randomness,stochastic network calculus theory needs to be extended to complex scheduling and network structure,which is still in research stage.In this thesis,the analysis scheme is designed from the perspective of modeling and measurement environment construction.The main research and contributions of this thesis are as follows:1.Modeling the arrival and service processes and extending the model to complex scenariosIn this thesis,network calculus is used as the model of arrival and service process.Considering that model could be affected in real environment,the model is modified and verified.For complex scenarios,the model is extended to different scheduling and multi-node scenarios.The theoretical accuracy of these models meets the requirements of performance analysis in complex network environment.2.The real network testing environment with high strength and high precision is realizedIn this thesis,a test environment with high performance and precision is built under the condition of general equipment,and multi-traffic scheduling and multi-node network structure testing can be realized.For different scenarios,the specific configuration of the software in the test is studied.Such as flow generator,flow capture filter,switch port instruction and scheduling configuration,The import method of data and fitting tools,other settings necessary for testing,etc.This effort improves the availability of flow test equipment.The test environment can verify the rationality of the modeling and the accuracy of the theoretical results3.Test results are compared with theoretical results in different scenarios.The network performance derived from complex network environment is tested,and the results of various theoretical derivations are compared.In various theoretical analysis results of different scenarios,the best algorithm of test results is found.The results can be used to maintain high accuracy and simplicity of theoretical analysis in different network environments.This thesis explores the application of stochastic network calculus in real networks.The results show that network calculus has strong adaptability and accuracy.It will become a powerful method to analyze network performance in the future.