Extending Kubernetes scheduler – analysis of available mechanisms

Abstract

The main goal of this paper is to analyze the component responsible for placing the pods onto appropriate nodes - the kube-scheduler, and the methods of extending it. The analysis is based on one of the available papers, that is [32], concerning implementation and using the custom scheduler, which goal was to minimize the resource contention phenomena. Based on it, there have been suggested some other methods of extending the vanilla scheduler, which gave similar results basing on different factors with much less effort. The analysis performed in this paper allowed to point out the pros and cons of each solution of extending the scheduler and indicate their use cases.