Research On NFV-based Resource Allocation In Data Centers

With the technology development of Cloud Computing,Internet of Things(Io T),Big Data and others,the scale of data centers is increasing continually.Traditional data center networks are faced with the problems of new business deployment,complex update,high energy consumption and low resource utilization.These problems pose great challenges to current data center networks.The emergence and development of NFV(Network Function Virtualization)technology reduces Capital Expenditures(CAPEX)and Operating Expense(OPEX)of data centers.NFV technology can help solve the problem of deploying new business and complex updating in traditional data center networks,and is a new direction for the research on data center energy consumption.The key of NFV technology is to decouple the network function from the proprietary equipment,and it implements the network functions in the form of software and deploys them onto the common hardware platform.SFC(Service Function Chain)including one VNF(Virtual Network Function)or more is an important application,which can provide flexible network service for users and has attracted a wide attention from both the academia and industry at home and abroad.This thesis mainly studies the SFC dynamic resource allocation in data center networks and the energy consumption of NFV-enabled data centers in inter-datacenter networks.At present,most researches focus on the problem of static SFC resource allocation under the NFV scenario,and many studies do not consider the types restriction of VNFs hosted in a server.To solve the problems above,this thesis establishes a mathematical model of the dynamic SFC resource allocation problem under the constraint of the types restriction of VNFs hosted in a server.Aiming at minimizing the deployment cost and end-to-end delay of a SFC,a heuristic algorithm for resource allocation is proposed in this thesis.In the proposed algorithm,the types of VNFs hosted in a node and the node computing resource of the data center network are comprehensively considered to complete the VNFs mapping based on hierarchical idea.Besides,the remained bandwidth of a links is considered to implement the virtual links mapping.The proposed algorithm is verified through simulation under the scenarios of both the Single-Feature Single-Tenant and Single-Feature Multi-Tenant.Simulation results show that the proposed algorithm can effectively reduce deployment cost and end-to-end delay of a SFC.The introduction of NFV technology in data centers can help reduce the OPEX in data centers.However,data center energy cost accounts for a significant proportion in the OPEX,and low resource utilization in data centers can also cause energy waste.Therefore,this thesis aims at reducing data center energy consumption and increasing data center resource utilization when studying the resource allocation problem of SFC.However,there is little research work to both study the SFC resource allocation problem and simultaneously consider reducing data center energy consumption in inter-datacenter networks.According to the energy consumption model of a server in the data center,this thesis firstly establishes the energy consumption model of the whole data center and then sets up the SFC resource allocation model for reducing data center energy consumption and end-to-end delay of a SFC.To solve the model,a genetic algorithm is proposed.In the proposed algorithm,the K shortest paths are firstly selected,and then the VNFs are deployed by genetic algorithm on the selected paths.The simulation results show that the proposed algorithm can effectively reduce end-to-end delay of a SFC,and to a certain extent improve data center resource utilization and reduce data center energy consumption.