A PBIL for delay constrained virtual network function placement with load balancing

This paper investigates the virtual network function placement (VNF-P) problem in network function virtualization (NFV), with load balancing and delay issues considered. In this problem, utilization of server and link resources is jointly minimized for load balancing purpose while end-to-end delay is constrained for better user experience. An integer-encoding population based incremental learning algorithm (IEPBIL) is presented to tackle the problem above, where an elitism preservation scheme and a mutation operation are devised. Simulation results show that IEPBIL outperforms a number of evolutionary and heuristic algorithms in terms of the success rate for placing virtual network functions.