Virtual Network Function Placement for Service Chaining by Relaxing Visit Order and Non-Loop Constraints

Network Function Virtualization (NFV) is a paradigm that virtualizes traditional network functions and instantiates Virtual Network Functions (VNFs) as software instances separate from hardware appliances. Service Chaining (SC), seen as one of the major NFV use cases, provides customized services to users by concatenating VNFs. A VNF placement model for SC that relaxes the visit order constraints of requested VNFs has been considered. Relaxing the VNF visit order constraints reduces the number of VNFs which need to be placed in the network. However, since the model does not permit any loop within an SC path, the efficiency of utilization of computation resources deteriorates in some topologies. This paper proposes a VNF placement model for SC which minimizes the cost for placing VNFs and utilizing link capacity while allowing both relaxation of VNF visit order constraints and configuration of SC paths including loops. The proposed model determines routes of requested SC paths, which can have loops, by introducing a logical layered network generated from an original physical network. This model is formulated as an Integer Linear Programming (ILP) problem. A heuristic algorithm is introduced for the case that the ILP problem is not tractable. Simulation results show that the proposed model provides SC paths with smaller cost compared to the conventional model.