GENESIS evolves three Deep Neural Networks (DNNs) using a Genetic Algorithm (GA) to optimally embed Service Function Chains (SFCs) on physical networks. Optimal SFC embedding consists of three sub-problems, namely optimal ordering of Virtual Network Functions (VNFs) in SFCs, optimally embedding VNFs on hosts, and optimally routing traffic between VNFS. These sub-problems have to be optimised simulatenously to achieve optimal SFC embedding. This is an NP-hard optimisation problem. I use a DNN to tackle each sub-problem and evolve these DNNs using a Genetic Algorithm to achieve optimal SFC embedding. GENESIS is 59% faster than the state-of-the-art GA while achieving 29% better performance.
Publications
Simultaneous Genetic Evolution of Neural Networks for Optimal SFC Embedding