Reseach On Tarffic Scheduling Application Of IP Network Over SDN

With the dramatic increase of users and traffic in the backbone network,Investment in backbone network paid by major operators is increasing, but the resource utilization rate of backbone network is still low. Traffic Engineering can solve this problem to a certain extent based on SDN architecture. Thus, it spurs the boom in research on existing IP backbone network traffic model and scheduling algorithm under SDN environment both in academic and industry. Among them, Google drives its data center across the WAN’s resouece utilization close to 100% through SDN technology. But, as the increasing size of the network, the number of traffic matrices is growing exponentially. Handling massive traffic matrices in traffic engineering affects the performance of scheduling algorithm and stability of network based on SDN architecture. So, it is unrealistic in network application analysis with a large number of trsffic matrices.Focused on issues related to low resource utilization rate of IP backbone network and unrealistic in handing massive traffic matrices in traffic engineering under SDN architecture, this paper firstly designed a new SDN application achitecture with reference to Coogle’s cross-data center WAN deployment scheme and the concept of critical traffic matrix. Then, this paper studied on the algorithm implement of critical traffic matrices getting module and routing computing module in the SDN application achitecture. Finally, this paper conducted a performance analysis of detail application using critical traffic matrices in traffic engineering.The details of this paper are as follows:(1) This paper designed a IP backbone application architecture under SDN based on critical traffic matrices. The main part of the application architecture is critical traffic matrices getting module, routing computing module and flow controlling module. This architecture intrduced the concept of critical traffic matrix baesd on SDN technology, aiming at just analysing critical traffic matrices for network applications, and saved the analysis results. it’s only need to call the analysis results of the maching critical traffic matrix, So as to achieve the purpose of improving the performance of traffic scheduling algorithm and stability of the network.(2) For the implement of critical traffic matrices getting module on SDN application architecture, this paper analyzed Critical-ness Acess Clustering(CritAC) algorithm and designed different clustering cost function as a solusion to the limitation of CritAC algorithm. This paper proposed some critacal matrices getting algorithm based on different cost function: Crital-ness and Time Acess Clustering(CritTimeAC) algorithm, RoutPolicy Acess Clusting(RoutPolicyAC) algorithm and Similarity-ness Aware Clusting(SimilarAC) algorithm, and analyzed the performance of these algorithms.(3) For the implement of routing computing module on SDN application architecture, this paper focused on the multi-constrained QoS routing problem and established a mixed integer programming, proposed a improved randomized solving algorithm based on the theory of A* algorithm to meet the QoS requirements and guaranteed to find a feasible path. Compared with existing algorithms and optimization model, the performance of this algorithm is more rapid and efficient.(4) In order to verify the effectiveness of the application architecture which mentioned in(1), this paper proposed route optimization scheme and multi-route switching scheme for multiple traffic matrices and discussed the performance of network vulnerability Analysis using critical traffic matrices instead of original traffic matrices.