A Survey of Autoscaling in Kubernetes

Abstract

Autoscaling is the vital feature of cloud infrastructure to acquire or allocate computing resources ondemand, which allows users to automatically scale the resources provisioned to the applications without human action under a fluctuating workload to optimize the resource cost while satisfying the Quality of Service (QoS) requirements. Kubernetes (K8s), the most prevalent container orchestration, provides built-in autoscalers to deal with the scaling problem in terms of vertical and horizontal at container level but still has some limitations. In this paper, we survey the state of the art of existing approaches to solve the problem of container autoscaling in Kubernetes, their main characteristics as well as their current issues. Based on the analysis, new future directions that can be explored are proposed.