Network function virtualization(NFV)architecture and software defined network(SDN)technology are being widely used to improve the rigid traditional network in the5 G era.NFV integrates virtual network functions(VNF)into high-capacity general-purpose servers,storage,and switches to separate software and dedicated hardware.A service function chain(SFC)is a series of VNFs arranged in a certain order and traversed by data traffic.An important issue in NFV environments is the deployment of SFC controlled by SDN technology,which makes network control efficiently programmable.It is critical to ensure quality of service(QoS)requirements of the request at a fine-grained level when a SFC is deployed.As the types and quantities of businesses on the physical network continue to increase,the service rate(bandwidth)allocated to the traffic through nodes becomes increasingly strained.Therefore,reasonable bandwidth allocation for virtual nodes in SFC can ensure the QoS requirements of delay-sensitive businesses,the efficient utilization of bandwidth resources,and improve network performance.This is the significance of this paper.The work and main contributions of this paper are as follows:(1)A fine-grained SFC deployment scheme is proposed and a hop-by-hop bandwidth allocation framework with QoS guarantee of statistical delay is established to guide SFC deployment.Based on the scheme and framework,two end-to-end tandem queuing systems with arrival processes of interrupted bernoulli process(IBP)and two-aggregation Markov modulated bernoulli process(2A-MMBP)are constructed for non real-time businesses and real-time businesses respectively.In order to evaluate the performance of SFC deployment,a new evaluation index is proposed:end-to-end packet average delay and delay violation probability,and a method is proposed to solve these two indexes.The effective bandwidth formula of complex Markov arrival process is derived.In order to realize the trade-off between the cost and the guarantee of delay QoS requirements,EB/EC theory is used to solve the service probabilities of nodes.And link bandwidth resources are allocated hop by hop.(2)A departure fitting algorithm is designed to fit the departure flow into IBP or MMBP,which is beneficial to the engineering implementation.Since the departure process is the arrival of the next hop,the analysis of the whole SFC tandem queuing system can be completed.An accurate expression of the departure process can be obtained by departure fitting,which is convenient to estimate the average delay and delay violation probability of each node.SFC deployment is formulated as an integer linear programming(ILP)problem with the goal of minimizing cost.To solve this problem,a new heuristic algorithm called hop-by-hop bandwidth allocation and SFC deployment(HBASD)is proposed.This paper compares the algorithm with other methods.Experimental results show that HBASD algorithm can obtain reasonable delay violation probability,meet the end-to-end delay requirements of heterogeneous businesses,and effectively save resources.(3)A more compact delay bound is obtained by analyzing SFC tandem queuing system in martingale domain.The arrival martingale processes of IBP,2A-MMBP and MMBP processes are constructed,and the service martingale process of geometric distribution is also constructed.The service probability of the service process at each node is derived hop by hop.Combined with the departure fitting algorithm,the hop-by-hop bandwidth allocation and deployment of SFC are completed under the guidance of martingale theory.Under the limitation of pre-distribution service,the two-phase network load minimization algorithm improved by Yen algorithm is adopted to obtain a better deployment scheme for heterogeneous businesses under the condition of meeting the delay requirements.Compared with HBASD algorithm,higher request acceptance rate can be obtained.This paper is the first to propose a hop-by-hop granularity bandwidth allocation framework for SFC under the delay QoS requirements of businesses.In this paper,the arrival processes of heterogeneous businesses are modeled as complex Markov processes.Under the guidance of effective bandwidth/effective capacity and martingale theories,different delay QoS requirements of heterogeneous businesses are guaranteed.Using departure fitting method,the precise departure process of traffic after serving can be obtained node by node,which is easier to be applied in engineering.The simulation results show the good performance of the proposed hop-by-hop bandwidth allocation and SFC deployment algorithm.The ideas and algorithms proposed in this paper have reference value for the future study of SFC deployment problems. |