Edge Serverless Autoscaling managed via Proximal Policy Optimization

Serverless computing is an emerging service model that is gaining consensus in both edge and core clouds. In fact, it allows developers focusing on application codes, breaking down services in multiple functions, without the burden of platform management. Services are run in edge clouds, closer to end-users’ devices, when service latency is an important constraint. However, latency control cannot be achieved with ample resource allocation, as in edge environments resources are limited and thus operational efficiency is essential. For these reasons, autoscaling of serverless functions has a central role, as it can lead to both reduced latency and efficient resource usage. In this work, we focus on the usage of reinforcement learning to drive resource-based autoscaling on OpenFaaS, the most-adopted open-source serverless platform, running on Kubernetes clusters. We use the Proximal Policy Optimization algorithm to dynamically configure the value of the Kubernetes Horizontal Pod Autoscaler, trained on Azure traffic traces. We discuss the algorithm configuration, focusing on state definition and relevant performance evaluation, including robustness with respect to traffic burstiness, as well as performance comparison with baseline approach and heuristics.