Research On Effective Placement Algorithms Of Service Function Chain

The technology of NFV has transformed the traditional service provision architecture into a software-based one.Traditional network functions are being substituted by VNFs which run on general hardwares.With the NFV-based network service framework,it is more challenging to guarantee the reliability of network services due to the fact that failure causes are more diversified.Redundancy is a common method to improve system reliability.Some existing redundancy method is coarse-grained and backups the entire SFC.Other redundancy methods protect the weakest node in an SFC.These methods neglect the shared physical resources brought by NFV and result in limited reliability improvement.In NFV-based edge computing network,VNFs are placed on the edge cloud serving low-latency and computing-intensive applications.The layerd structure of edge computing network and the latency-sensitive nature of its received service requests bring challenges to the service offloading problem of the SFC.Existing VNF placement problems in edge computing network rely on centralized optimization methods.These work lack fairness consideration for individual applications.To tackle aforementioned problems,this thesis has the following contributions:1)To solve the over backup and resource under-utilization problems in the existing methods,we design a novel redundency placement algorithm CERA.In CERA,CIM is used to direct the backup selection process.Backup VNFs are then iteratively added to the SFC by the assorted update process.We prove the effectiveness of CIM through detailed mathematical deduction.The simulations demonstrate that compared with the benchmark algorithms,CERA improves the cost-efficiency while guaranteeing the reliability of SFCs.2)We design a distributed SFC placement method to solve the SFC placement problem in the NFV-based edge computing network.We first model the problem as a potential game with dedicated potential function.By exploiting the properties of the potential game,we design the DSOA algorithm to obtain the game's nash equilibrium.The convergence and Po A(the Price of Anarchy)of the algorithm are proven by mathematical deductions and demonstrated by simulations.Simulations also show that DSOA can guarantee better fairness of SFCs with approximate results in comparison with the Cross-Entropy Method.