Service Function Chain Scheduling And Virtual Network Function Placement In Datacenter Environment

With the development of ICTs(Information and Communication Technologies),people are more and more addicted to the network technologies,through which we can acquire knowledge and have entertainment.Moreover,with the explosive emerging of new network services,such as short/live video,online shopping,smart IoT etc.,higher network volume,shorter service latency,and higher service availability are claimed on current networks.Thus many new network technologies are developed to meet these increasing QoS requirements.In the area of wireless communications,5G is ready for the market.In the core of mobile networks,cloud-based service model has been a main trend,which greatly promotes the applications of cloud computing and edge computing.Network sys-tems need more power and flexibilities on the automated management,resource detec-tion and allocation etc.,while keeping the capital expenditures and operating expenses(CAPEX/OPEX)at a lower level.All these requirements make the virtualization tech-nology one of the keystones of the future networks.Network function virtualization(NFV)allows network functions(NFs)or mid-dleboxes traditionally attached on specific hardwares to be realized in software and dynamically outsourced to be run on any common off-the-shelf servers,which brings great conveniences and flexibilities in programmability,management and policy inter-actions between virtual NFs(VNFs).Nowadays,more and more network operators and enterprises leverage NFV technology to reduce the cost of infrastructure construction and management.In NFV enabled networks,users' service requests are accomplished by service function chains(SFCs),each of which is composed of a series of ordered VNFs.Thus users' requests can also be called SFCRs(SFC Requests).Given a series of SFCRs,we should map them into the networks and deploy related VNFs to serve them.VNFs can be instantiated and migrated among servers flexibly,which gives the network operators opportunities to utilize the network resource in a more efficient way.NFV will be widely used in the future networks,and mostly in datacenter environ-ment.So how to utilize the network resources in an optimal way given a set of SFCRs while meet the QoS requirements of different services,which is also known as the SFC scheduling and VNF deployment/placement problem,should be well resolved.This dissertation concentrates on the SFC scheduling and VNF placement in dat-acenter environment.According to different QoS requirements,the research can be divided into following three parts:· 1:SFC scheduling and VNF placement in cloud datacenter considering time-varying workloads.· 2:SFC scheduling and VNF placement in cloud datacenter considering service availability.· 3:SFC scheduling and VNF placement among distributed datacenters.Our major contributions can be summarized as:1:To improve the utilization efficiency of network resources,two less-considered factors are taken into consideration,which are the time-varying workloads,and the basic resource consumption(BRC)when instantiating a VNF.Then the SFC scheduling and VNF placement problem is formulated as a binary integer linear programming(BILP)model with the aim of minimizing the number of used servers.Afterwards,a Two-StAge heurisTic solution(T-SAT)is designed to solve the BILP.T-SAT consists of a correlation-based greedy algorithm for SFCR mapping(first stage)and a further ad-justment algorithm for VNFRs(VNF Requests)in each SFCR(second stage).Finally,T-SAT is evaluated with the artificial data generated according to Gaussian function and trace data derived from Google's datacenters.The simulation results demonstrate that the number of used servers derived by T-SAT is near to the optimal results and much smaller than the benchmarks.Besides,it improves the network resource utilization sig-nificantly.2:Although NFV brings great cost reduction and operation flexibility,the inad-equacies that are inherent in it should not be ignored.One of the most noteworthy factors is the availability of network services,owing to the error prone nature of soft-ware.The SFC scheduling and VNF placement considerin g availability has attracted attention in academics,and reserving redundancy has been seen as the de facto technol-ogy to improve availability.Compared with the traditional backup schemes for physical machines,the resource orchestration of software is more flexible,so the characteristics of software should be taken advantage of to improve the resource efficiency.To im-prove the resource utilization efficiency further,multi-tenancy technology and sharing mechanism of redundancy are considered.Then the problem is formulated mathemat-ically,and a joint deployment and backup scheme(JDBS)is proposed.To evaluate the performance of JDBS,a numerical simulation is made in detail and JDBS is com-pared with 4 contrasting schemes in existing literatures.The simulation results show that JDBS has a great advantage over the contrasting schemes and can save about 40%resources at most.3:The research is studied in edge computing environment,which is a rising dis-tributed cloud computing architecture.NFV and edge computing are two promising and innovative technologies to accelerate 5G networks.However,placing the SFCs into the edge computing enabled networks is an intractable issue and some new challenges shall arise.Firstly,edge computing is a hierarchical and geo-distributed structure,which will influence the form of SFC and make the VNF placement location-related.Secondly,the data processing in edge computing is hierarchical too,which incurs different latency re-quirements.Multi-tenancy technology and chaining of VNFs are also considered,which will raise the need of tradeoff between node resource consumption and bandwidth con-sumption when placing VNFs.Apart from the above challenges,the balancing about the utilizing of different types of resources is considered,trying to reduce the "resource bottlenecks" in the networks.Then the above problem is formulated as a BILP model mathematically aiming to minimize the total resource consumption,which is proven to be NP-hard.The optimal results is derived when the number of SFCRs is small taking advantage of optimization solver and a polynomial time heuristic is proposed when the problem scale is large.Simulation results show that the resource consumption derived by the proposed heuristic solution is near to the optimal solution and its performance is very much superior to the contrastive schemes too.