Simultaneous Genetic Evolution of Neural Networks for Optimal SFC Embedding

TL;DR

This work tackles the NP-hard Optimal Service Function Chain Embedding problem by proposing GENESIS, a genetic-evolution framework that simultaneously trains three sine-activated neural networks to drive CC, VE, and LE solvers. By encoding solution candidates as gradients of NN hidden layers, and using a Gaussian sampler for VNF embedding and an A* path planner for LE, GENESIS achieves global optimization with a compact, scalable encoding. Empirical results across 48 data-centre scenarios show GENESIS converges to optimal solutions in all cases and outpaces competing GAs in both quality and speed, while a greedy baseline lags in both respects. The approach advances practical OSE by delivering reliable, scalable performance suitable for dynamic, large-scale networks.

Abstract

The reliance of organisations on computer networks is enabled by network programmability, which is typically achieved through Service Function Chaining. These chains virtualise network functions, link them, and programmatically embed them on networking infrastructure. Optimal embedding of Service Function Chains is an NP-hard problem, with three sub-problems, chain composition, virtual network function embedding, and link embedding, that have to be optimised simultaneously, rather than sequentially, for optimal results. Genetic Algorithms have been employed for this, but existing approaches either do not optimise all three sub-problems or do not optimise all three sub-problems simultaneously. We propose a Genetic Algorithm-based approach called GENESIS, which evolves three sine-function-activated Neural Networks, and funnels their output to a Gaussian distribution and an A* algorithm to optimise all three sub-problems simultaneously. We evaluate GENESIS on an emulator across 48 different data centre scenarios and compare its performance to two state-of-the-art Genetic Algorithms and one greedy algorithm. GENESIS produces an optimal solution for 100% of the scenarios, whereas the second-best method optimises only 71% of the scenarios. Moreover, GENESIS is the fastest among all Genetic Algorithms, averaging 15.84 minutes, compared to an average of 38.62 minutes for the second-best Genetic Algorithm.

Figures (7)

• Figure 1: In implicit Link Embedding optimisations, studies assume that there are direct physical paths between hosts, ignoring the network switches. In explicit Link Embedding optimisations, studies consider the network switches and the possibility of multiple paths connecting hosts.
• Figure 2: The evolver uses the gradients of three NNs as the genetic encoding of the GA to optimise the OSE sub-problems simultaneously. Three solvers use the output of the NNs to decode the encoding and generate a deployable candidate solution. The process of decoding is sequential; however, the gradients are optimised simultaneously.
• Figure 3: The predictors are fully connected, feed-forward DNNs with one hidden layer consisting of two neurons. The neurons use the sine activation function to increase exploration.
• Figure 4: A DNN with gradients specified. V1's gradients are greater than V2's.
• Figure 5: A diagram depicting the process of translating genetic encoding ($w1$ and $w2$) to decide the host to embed a VNF on.