Intelligent Network Service Embedding using Genetic Algorithms

Network Function Virtualization (NFV) opens us great opportunities for network processing with higher resource efficiency and flexibility. Nevertheless, intelligent orchestration mechanisms are required, such that NFV can exploit its potential and fill up to its promise. In this respect, we investigate the potential gains of embracing Artificial Intelligence (AI) for the virtual network function (VNF) placement problem. To this end, we design and evaluate a genetic algorithm, which seeks efficient embeddings with runtimes on par with heuristic methods. Our proposed embedding method exhibits innovations in terms of network representation and algorithm design, thereby, deviating from typical genetic algorithms. Compared to a heuristic, the proposed genetic algorithm yields higher request acceptance rates, stemming from more efficient resource utilization. We further study a range of factors and parameters that affect the efficiency of the genetic algorithm.