Traffic Matrix Estimation And Network Traffic Scheduling In Software Defined Networking

Traffic engineering(TE),a typical application in software defined networking(SDN),mainly studies the measurement and management of network traffic,and it improves network resource utilization and guarantees quality of service(Qo S)by designing feasible routing mechanisms to schedule network traffic.However,TE based on traditional network technology is limited with the continuous expansion and application of cloud computing and Internet of Things.One the one hand,the expansion of network scale makes it difficult to deploy new protocols in closed network equipment,which increases the difficulty for operators to customize network services.On the other hand,the increase in user demand and the emergence of various new services have increased the cost of operation and maintenance,as well as the difficulty of network function management.SDN and network function virtualization(NFV)technology are the keys to the efficient operation of data center networks(DCNs)with core functions such as big data computing,storage,and analysis.Compared with traditional networks,the characteristics of decoupling of control and data planes in SDN as well as separation of software and hardware in NFV have huge advantages for supporting TE.Although SDN has expanded the scope of network resources,bandwidth and flow tables have become limited network resources,which puts forward higher requirements for accurate network traffic measurement,and efficient and flexible scheduling of network resources and network functions.With the development of SDN and NFV,it is of great practical significance and application value to carry out research on SDN-based TE.Combining with the latest research in this field,this thesis mainly studies the problem of TM estimation and network traffic scheduling in SDN.The main work and contributions are summarized as follows:1.TM estimation through large size flow identification in SDN.The hybrid network monitoring scheme is obtained by combining partial direct measurement offered by SDN with some inference techniques.Specifically,gradient boosting machine(GBM)is first used to identify the large size flows from multiple historical TMs,and the sampled origin to destination pair(OD)pairs whose flow size is selected as large size flow are found.Then,a greedy heuristic algorithm is proposed to solve SDN-enabled switch selection problem,so as to further best utilize the flow table resources and guarantee that most of sampled OD pairs are measured in the flow table.A source node prefix tree based bit merge aggregation(SPTBMA)is also proposed to design feasible forwarding rules and reserve more flow table resources for sampled OD pairs.Finally,the experimental results based on real traffic dataset demonstrate that the proposed scheme outperforms existing algorithms in terms of improving TM estimation accuracy and overcoming limitation of flow table resources.2.Joint traffic-aware consolidated middleboxes selection and routing in distributed SDNs.In SDN with static configuration mechanism,dynamic changes in network traffic will not only affect the link load in the data plane,but also the load among controllers in the control plane.Together with SDN,NFV brings flexibility into software middleboxes management,but different network functions in middleboxes may alter the volume of processed traffic,leading to high congestion occurred in specific bottleneck links if middleboxes selection and traffic routing are not well jointly planned.To ensure the quality of service(Qo S)in both control and data planes,this problem is first formulated as a traffic-aware consolidated middleboxes selection and routing(JTMSR)problem.Then,a two-phase RL＿RFRD algorithm is designed to solve this NP-hard problem where the first phase is to redirect selected flows by applying wildcard rules and the second phase is to find fine-grained routing path by a rounding-based algorithm.Finally,the experimental results show that the proposed method has near-optimal controller load balancing and link load balancing performance and can reduce response time by 2-5 times.3.Service function chain(SFC) reconfiguration for network slicing in 5G.Since more and more users are frequently trying to access their customized service in 5G networks,SFC requests are expected to be reconfigured dynamically and adaptively according to the varying traffic demand and available resource.However,SFC reconfiguration involves flow-rerouting,virtual network function(VNF)instances scaling and migration,which consumes additional resource and leads to service interruption,further degrading user experience.In order to solve this problem,the maximum acceptance ratio minimum reconfiguration cost optimization problem is formulated by considering existing resource and scaling resource.The objective within existing resource is to perform reconfiguration with minimum reconfiguration overhead while make profit by admitting requests with positive revenue when resource is scaled.Then,a low-complexity heuristic algorithm is designed to solve this NP-hard problem.Finally,the experimental results show that the proposed algorithm can significantly ensure the accept ratio of requests and reduce reconfiguration cost.