An Auto-scaling Framework for Containerized Elastic Applications

Mechanisms for automatic resource provisioning in cloud environment have been widely studied. However, most current work are based on Virtual Machines(VMs) and few auto-scaling systems dedicated for containerized environment have been presented. This paper proposes an auto-scaler that caters for containerized elastic applications. We devise a hybrid scaling strategy based on a resource demand prediction model to meet Service Level Agreements(SLAs) in face of quickly varying workloads. We implemented our system in a private container-centric cloud environment and experimental results demonstrate that our auto-scaling framework outperforms existing rule-based approaches.