A Study Of Service Function Chain Assurance Based On Software Defined Technology

With the increase of Internet applications,users need to obtain more network services to meet application requirements.Network operators often provide users with network services through specialized network devices,namely middlebox.Load balancers,QoS Monitors,Video Transcoders,Gateways,and Proxies are all examples of the middleboxes.However,providing services by hardware middleboxes has disadvantages such as high price,complex management,and poor flexibility.Network Function Virtualization(NFV)technology came into being.Using software to realize network function,namely virtual network function(VNF),NFV achieves low-cost,highly flexible,and modular middlebox processing.Network services require flows to pass through a set of middleboxes(or VNFs)in a specific order,known as a service function chain(SFC).Operators deploy multiple middleboxes in the service function chain on different servers to provide users with flexible end-to-end network services.Meanwhile,NFV technology can reasonably allocate computing and storage resources for the middleboxes of the service function chain,which effectively improves server resource utilization and reduces operating costs.The existing service function chain assurance solutions have the problems of high operation cost,significant resource overhead,poor robustness and flexibility,and increased network delay,which significantly reduces the quality of services.This dissertation studies the service function chain assurance problem to solve the above challenges,combined with the advantages of centralized management and flexible scheduling of softwaredefined technology.Specifically,this dissertation considers numerous devices,massive traffic,diverse needs,limited resources,and network dynamics as the research foothold,aiming to provide efficient service function chain assurance,focusing on four aspects:function deployment,service schedule,traffic routing,and network update.The main research contents and contributions include:Joint Server and VNF Incremental Deployment.Function deployment is the premise of service function chain assurance.Appropriate network function deployment can effectively reduce the network function and link load.Our research is based on the following facts and observations.First,the server may not have been deployed,and VNF deployment cannot be achieved.Second,some VNFs(such as load balancers)only need to process part of flows to obtain high network performance.However,existing work often assumes that each type of VNF needs to process all flows,resulting in the excessive number of servers used and increasing network operating costs.To this end,we propose a joint server and VNF incremental deployment problem and prove that the problem is NP-hard.We design an approximate algorithm based on the greedy knapsack problem,which can achieve 2·H(q·d)approximate ratio,where H is the harmonic function,q is the number of VNF types,and d is the maximum number of flows passing through the switch.The experimental results show that the proposed algorithm only needs to deploy a near-optimal number of servers to meet the VNF requirements.Robust and Flexible Network Service Scheduling.Service scheduling refers to the selection of required middleboxes for traffic.Flexible service scheduling can reduce network function load and improve network throughput.Existing works often focus on meeting resource constraints and connection consistency while ignoring the requirements of network robustness and flexibility,resulting in reducing network service quality.Therefore,this dissertation designs a robust and flexible service scheduling(RFSS)system.We implement a bucket-based forwarding scheme using group entries to ensure connection consistency in the data plane.In the control plane,we propose a rounding-based bucket allocation algorithm to enhance the robustness and flexibility of the network.This algorithm can realize the bi-criteria constant approximation.The simulation results show that compared with other schemes,the RFSS system can improve the network throughput by about 150%while meeting the robustness and flexibility requirements.Mobility-Aware Service Function Chain Routing.Traffic routing is the core of service function chain assurance,and efficient routing can reduce resource overhead and transmission delay.User mobility is an important challenge for service function chain routing.Existing works often need to deploy multiple paths for each request,resulting in high flow entry overhead.Therefore,this dissertation designs an efficient routing scheme called MASCOT,which realizes service function chain routing through location prediction,path decision,and data forwarding.We adopt the k-order Markov method to predict the base station that the user will visit next for location prediction.For routing decision,we propose the service function chain routing(SRS)problem,design an online service function chain routing algorithm based on the primal-dual method,and prove that the algorithm can achieve a good competition ratio.We propose a flow table installation scheme based on segment routing for data forwarding,which reduces the flow entry overhead.Extensive simulation results show that compared with the state-of-the-art methods,the MASCOT scheme can improve the network throughput by about 40%.Real-Time Service Function Chain Routing Update.Network update is a supplement to the service function chain assurance.Real-time network updates can provide stable network services and ensure user service quality.Due to network dynamics,the previous routing configuration may not suit the current network environment or even cause network congestion,so a network update is required.Most existing solutions determine new network configurations based on the current network load,then update VNF deployment and flow routing.However,large-scale VNF state migrations and many flow entry installations can cause long update delays.This dissertation studies a real-time service function chain routing update problem to address the challenges of flow state consistency and update latency.We propose a flow state parallel update scheme in the data plane,significantly reducing the cache overhead and the state update latency.We propose an NFV network update problem in the control plane that satisfies the delay requirements to determine the target configuration.Then,we prove that the problem is NP-hard and propose an algorithm with an approximate ratio of 3logn/α+3 based on the random rounding method,where n is the number of middleboxes and a is a constant related to the middlebox processing capacity.To determine the update order of requests,we propose a greedy mechanism-based routing and scheduling algorithm comprehensively considering the constraints of link bandwidth,controller.cache capacity,and middlebox processing capacity to achieve real-time congestion-free update scheduling.The experimental results show that our method can reduce the network update delay by about 86%compared with the previous methods,while the VNF instance load ratio increases by less than 5%.