KAPETÁNIOS: Automated Kubernetes Adaptation through a Digital Twin

Abstract

This demo presents a self-operating Kubernetes (K8s) cluster that uses digital twinning and machine learning to autonomously adapt its Horizontal Pod Autoscaler (HPA) to workload changes. The demo uses a digital twin of a K8s cluster to gather performance statistics and learn a model for the workload. With the model, the cluster autonomously adjusts HPA parameters for better performance. The demo illustrates this process and shows that the requested pod seconds decrease by ∼37 %, while the request latency stays mostly unaffected.

HyPA: Hybrid Horizontal Pod Autoscaling with Automated Model Updates

“…In this section, we evaluate HYPA and compare it against a range of baselines. All presented results are obtained from the discrete event-based simulator presented in [17]. It combines data-driven models as well as white-box re-implementations of k8s' components.…”

“…2) Proactive: The proactive component uses data-driven models to forecast the expected number of requests in each category for the next time slot. Based on the estimated number of requests and a model of the work that each request causes, the so-called application profile [17], the proactive component estimates the number of replicas in the system:…”