Research On Design And Embedding Of Virtualized Network Function Forwarding Graph

Because of the traditional network functions coupled with existing network topology too closely,the deployment of traditional service function chains(SFC)is complex,with poor flexibility and scalability.As a key technology to solve this problem,Network Function Virtualization(NFV)has received wide attention from the industry since it was put forward.It is possible to centrally manage and flexibly configure service function chains,and expand resources dynamically,which is able to shorten the deployment time of SFCs.In NFV,Virtualized Network Function Forwarding Graph(VNF-FG)could be used to describe a logical topology containing multiple service function chains.But operators are faced with the challenge of maximizing their own interests while satisfying the quality of users' service requests.Therefore,aiming at this service-oriented resource management problem,we conduct research on VNF-FG designing and embedding in this thesis.For the VNF-FG designing problem,this thesis divides it into two phases: the first step is to design a single SFC,and the second is to form a VNF-FG by aggregating multiple SFCs.(1)For the designing of single SFC,TCRSFCD is proposed,aiming at minimizing the overall consumption of logical link bandwidth.TCRSFCD guarantees VNF dependencies in network service request,and reduces the bandwidth of subsequent links,by placing the VNF with smaller ratio of traffic change in the front position of SFC.(2)For the aggregation of multiple SFCs,two algorithms,FMAA and LMAA,are proposed,aiming at minimizing the number of VNF node.Furthermore,we define a new metric called Aggregation Degree(AD)to quantize the degree of aggregation on a certain path in VNF-FG.These two multi-SFCs aggregation algorithms group SFCs with sourcedestination endpoint pair,and then aggregate SFCs within the group,with AD as a reference index.And the simulation results show that TCRSFCD is superior in overall logical link bandwidth consumption,resource requirements of VNFs and maximum logical link bandwidth requirement.Compared with FMAA,LMAA can reduce the number of nodes in VNF-FG to a greater extent on the basis of minimizing link bandwidth consumption.For the problem of VNF-FG embedding,a Traffic-Aware based Forwarding Graph Embedding(TA-FGE)algorithm is proposed in the off-line scenario,with the goal of minimizing the embedding cost.TA-FGE uses K-shortest paths(KSP)as the embedding paths in substrate network,and the Difference Factor defined in this thesis is used to determine the embedding order of nodes in VNF-FG.TA-FGE makes the link bandwidth consumption as small as possible,so as to achieve the purpose of saving embedding cost.Finally,the simulation results show that TA-FGE performs well in reducing embedding costs,and the success rate of embedding can also be ensured by setting an appropriate value K of KSP.