A Composition Selection Mechanism for Chaining and Placement of Virtual Network Functions

Virtual Network Functions (VNF) reduce the complexity to deploy a Network Service (NS), yielding flexibility, and scalability to attend new market appliances and minimizing the required investments. An ordered set of VNFs to serve a NS is called a Service Function Chain (SFC). Most of the literature works deal with SFCs generation and embedding as independent problems, not taking into account the residual substrate network (SN) status during the SFCs generation stage. Furthermore, a NS can award some flexibility regarding the VNFs sequence, i.e., only a subset of VNFs have a fixed precedence requirement. Following this idea, the network operator may enjoy some options to serve a NS. In this work, we introduce a new embedding approach which selects the SFC composition that best fits the residual SN, leading to a better performance in the chaining and placement of VNFs. The results showed that the composition selection mechanism increases performance compared to traditional models that use a fixed composition, improving resources sharing, and increasing the network operator revenues. Also, we demonstrated that the generation of an optimal SFC for a NS does not always lead to the best acceptance rates in a network with partially consumed resources.