Cluster autoscaler for unmanaged Kubernetes cluster development on cloud

Abstract

This project report comprises details of the research "Cluster Autoscaler for Unmanaged Kubernetes Cluster Deployment on Cloud". Underutilization of server resources is a huge issue in enterprise data centers. When it comes to Kubernetes, underutilization and overutilization issue exists as is. In Kubernetes main course of having this issue is the use of fixed number of Kubernetes worker nodes. The Kubernetes community provides a cluster autoscaler solution to reduce underutilization and overutilization on Kubernetes clusters. This solution is only supported by a few major cloud providers like Google, AWS, DigitalOcean, and few others. Also, this solution is tightly bound to the auto scale group concept in those clouds. Hence this solution provided by the Kubernetes community cannot be used elsewhere. Therefore, there is a necessity for a general auto scaling approach that can be used on a wide range of cloud platforms and hardware virtualization platform. This research is to design and develop a Kubernetes Cluster Autoscaler which can be used on any cloud platform. This is achieved by removing the tightly bound auto scale group in the solution proposed by this research. Proposed solution use API and SDK provided by cloud provider and using libvirt which is a general purpose API library to manage KVM, Xen, VMWare ESXi and QEMU.