Design And Implementation Of FPGA-based Intelligent Traffic Control System For Three-layer Networks

With the rapid growth of network applications in recent years, the departments,such as government, education and economic, have established large and very largelocal area networks. Many of these networks have typical three-layer hierarchicalstructure, thus they can be referred as three-layer networks. In the three-layer network,the users’ population is increasing, and new applications are emerging. Improvingnetwork performance, optimizing network configuration, and guaranteeing the criticalbusiness traffic of the network have become challenge of three-layer networks.Because the existing traffic control methods cannot be directly applied on the endnodes of the three-layer network, and cannot meet the traffic request of individual usersin the access layer of the network, by introducing the genetic algorithm and improvingthe performance of existing network inspection methods, based on the pros of FPGAsuch as fast configuration and reprogramming, an intelligent traffic controller system forthree-layer network is designed and implemented in this thesis.Firstly, based on the analysis of advantages and disadvantages of three networkservice inspection technologies, a new inspection method which combines DPI (DeepPacket Inspection) and port mapping inspection technologies is proposed. In the method,consulting the Snort intrusion detection system, in order to improve the inspectionefficiency, inspection rule sets are reconstructed by adding the application rules in thethree-layer network and forming state set consisted of extracted key features, dualmulti-pattern matching and filtering mechanism is implemented by multilevel linked list,and traditional application is recognized by port mapping inspection method. Theexperiment result shows that this method has higher recognition rate and the averagenumber of rules can be decreased to26%-86%of its original number.Secondly, by analyzing the impact of non-converge prematurity problem of thegenetic algorithm on intelligent control, adaptive improvement of individual sub-groupsis carried out, and thus an adaptive hybrid genetic algorithm is designed. There areimprovements on adaptive crossover and mutation operator, on selection operator byintroducing artificial immune algorithm, and on fitness function by penalty factors insimulated annealing algorithm. The experiment result shows that the improvedalgorithm can effectively restrain the prematurity, improve the population diversityduring later evolution period, activating co-evolution among groups, and increase thesearching capability, and thus weaken the impact of genetic premature convergenceproblem on intelligent control.Finally, a FPGA based intelligent traffic controller system for three-layer networkis designed and implemented. The system consists of traffic identification sub-systembased on proposed inspection methods and traffic identification sub-system on the basis of adaptive hybrid genetic algorithm. The traffic identification sub-system is designedunder the guiding of modular methodology, and its functional models are designed,coded, synthesized, simulated and verified in the integrated environment of XilinxISE14.2. After that, circuit diagram is given.