| In recent years,with the increasing diversification of Internet busniess,people show a higher demand for the network services.To meet this growing demand,network service providers need to deploy more network functions(NFs)in the ISP network,and provide users with the required network services through their combination.In traditional scenario,network functions are implemented by dedicated hardware appliances.These middle-boxes are generally expensive,vendor-specific and location fixed which makes it difficult to dynamically scale their capabilities and add new functions to the existing devices.To solve the dilemma faced by traditional networks,a novel network architecture framework named Network Function Virtualization(NFV)has been proposed as a promising technology for the future network service providers.Through hardware and software decoupling,NFV enables to replace the dedicated devices with software instacnes running on commodity servers and realize the flexible deployment and removal of NFs.In NFV scenario,network serice flows need to traverse through a series of Virtual Network Functions(VNFs)to get a complete end-to-end service from source to destination,which is known as VNF-Service Chain(VNF-SC).In the era of big data with high throughput and high concurrency,Elastic Optical Network(EON)provides a reliable underlying support for VNF-SC deployment with its huge bandwidth capacity and flexible spectral allocation.However,this flexibility also brings new cahllenges to the management of multi-domain network resources.This paper mainly studies how to efficiently allocate network resources when deploying VNF-SC in Elastic Optical Network.Firstly,we propose the concept of Resource-virtualized Elastic Optical network(Rv EON)based on the existing EON architecture,and study the joint optimization problem of multidomain network resources when deploying VNF-SC in Rv EON.We formulate an integer linear programming(ILP)model to solve the problem exactly,and then several algorithms are proposed,including a Maximum Reuse with Decentralization Principle(MR-DP)based heuristic algorithm.Simulation results indicate that when we reuse the deployed VNF instances as much as possible,it can not only effectively improve the IT resource utilization,but also bring some improvement to the utilization of other resources in the network,and achieve the goal of joint optimization of multi-domain resources.Secondly,we consider the impact of VNF reuse on processing delay of each node,and study the delay-aware and resource-efficient deployment of VNF-SC in EON.To solve this problem,we also formulate an ILP model and design a Delay-Aware and Load-Balancing Mapping Algorithm(DALB-MA)to solve the high complexity problem of the mathematical model.The main design mentality of DALB-MA is that when the number of requests are small,we try to reuse the deployed VNF instacnes to fully improve the utilization of IT resources.When the number of requests increases,we consider instantiating more VNF instances,exchanging resources for time,to limit the processing delay to a certain range.Simulation results indicate that the proposed heuristic can simultaneously improve the utilization of IT resources and spectrum resources,and limit the processing delay to a certain range.Compared with the benchmark algorithms,it greatly reduces the blocking rate of requests,improves the performance of hthe network,and realizes the joint optimal allocation of multi-domain network resources. |
PDF Full Text Request