On Multicast-Oriented Virtual Network Function Placement: A Modified Genetic Algorithm

摘要

Network function virtualization (NFV) is an emerging network paradigm that will ease the network reconfiguration and evolution for Network Service Providers (NSPs). In NFV, the virtual network function placement (VNFP) problem has become a hot topic. However, little research attention has been paid to multicast-oriented VNFP (MVNFP) problem. This paper studies the MVNFP problem and presents a two-step approach to address it. The first step constructs a multicast tree for a given multicast service request and the second one places VNFs onto the tree. In the first step, Dijkstra's algorithm is adopted while in the second step, a modified genetic algorithm (mGA) with problem-specific chromosome encoding, crossover and mutation is proposed. Simulation results show that mGA performs better than a number of evolutionary algorithms with respect to the solution quality and convergence.