Commodore: Fail Safe Container Scheduling in Kubernetes

Abstract

Kubernetes is a tool to facilitate deployment of multiple virtualized applications using container technology. Kubernetes scheduling mechanism orchestrates computing resources per application at runtime. However, resource allocation is static, as the maximum amount of computing resources that each application can use, must be reserved in advance. If the application requests more resources than the maximum, a fail scheduling event is generated. Although solutions to the problem of automatic scaling in Kubernetes are known to exist and automatic scaling is supported by cloud providers such as Amazon and Google, these solutions are fully proprietary and not generic (e.g. do not apply to all Kubernetes distributions). Our solution, referred to as “Commodore”, is capable of allocating (or de-allocating) resources based on the actual demands of running applications. Taking advantage of the virtualization features of cloud computing, applications are deployed on worker machines (nodes) as Virtual Machines (VMs). This not only results in better utilization of computing resources (i.e. CPU, memory and network are defined virtually) but also, in enhanced software security by isolating services or applications from each other. The experimental results demonstrated that Commodore responds to the increasing (or decreasing) resource demands of each application leading to significantly faster response times compared to a non-auto scaled implementation.