Service Function Chains (SFCs) enable programmatic control of the functions and services in a computer network. By leveraging Software Defined Networking to control the links between virtualised network functions, SFCs provide a scalable approach to dealing with the increased pressures on network operation and management. However, embedding SFCs onto the underlying physical network and compute infrastructure is an NP-hard problem. Genetic Algorithms (GAs) have been used to address this issue, but they require significant time to evaluate solution quality (fitness) online, with most existing approaches instead adopting offline simulations or analytical evaluations. To enable online use of GAs in solving the SFC embedding problem, we introduce a hybrid online-offline approach to efficiently evaluate the fitness of generated solutions. At the core of this is BENNS: a surrogate model that approximates fitness and is agnostic to topology, traffic, and SFC-embedding. We evaluate our approach in a static environment across five experiments, varying available resources and traffic loads, and in a dynamic network environment. Our results demonstrate that our approach is capable of exploring thousands of potential configurations and generating deployable solutions in 19.1 minutes on average, compared to online-only approaches, which take 17.8 hours on average to explore ten solutions in our experiments and do not converge on an optimal solution.

翻译：服务功能链（SFCs）实现了对计算机网络中功能与服务的可编程控制。通过利用软件定义网络来控制虚拟化网络功能间的链路，SFCs为应对网络运营与管理日益增长的压力提供了一种可扩展的解决方案。然而，将SFCs嵌入到底层物理网络与计算基础设施是一个NP难问题。遗传算法（GAs）已被用于解决此问题，但其在线评估解质量（适应度）需要大量时间，现有方法大多转而采用离线仿真或解析评估。为使遗传算法能在线应用于解决SFC嵌入问题，我们提出了一种混合在线-离线方法，以高效评估生成解的适应度。其核心是BENNS：一种对拓扑、流量及SFC嵌入方式均保持无关性的代理模型，用于近似估计适应度。我们在静态环境中通过五项实验（改变可用资源与流量负载）及动态网络环境中评估了该方法。实验结果表明，相较于纯在线方法（在我们的实验中平均需17.8小时探索十个解且未收敛至最优解），我们的方法能够探索数千种潜在配置，并平均在19.1分钟内生成可部署的解决方案。