Kubernetes cluster optimization using hybrid shared-state scheduling framework

This paper presents a novel approach for scheduling the workloads in a Kubernetes cluster, which are sometimes unequally distributed across the environment or deal with fluctuations in terms of resources utilization. Our proposal looks at a hybrid shared-state scheduling framework model that delegates most of the tasks to the distributed scheduling agents and has a scheduling correction function that mainly processes the unscheduled and unprioritized tasks. The scheduling decisions are made based on the entire cluster state which is synchronized and periodically updated by a master-state agent. By preserving the Kubernetes objects and concepts, we analyzed the proposed scheduler behavior under different scenarios, for instance we tested the failover/recovery behavior in a deployed Kubernetes cluster. Moreover, our findings show that in situations like collocation interference or priority preemption, other centralized scheduling frameworks integrated with the Kubernetes system might not perform accordingly due to high latency derived from the calculation of load spreading. In a wider context of the existing scheduling frameworks for container clusters, the distributed models lack of visibility at an upper-level scheduler might generate conflicting job placements. Therefore, our proposed scheduler encompasses the functionality of both centralized and distributed frameworks. By employing a synchronized cluster state, we ensure an optimal scheduling mechanism for the resources utilization.