Optimizing Cloud Service Efficiency with Infrastructure-Aware Scheduling

Abstract

With the rapid development of cloud-native computing technologies, containers have become an integral part of cloud infrastructure. At the same time, Kubernetes has positioned itself as a solid foundation for cloud-native orchestration platforms. However, the exponential growth of the container adoption raises a concern over energy usage. Existing Kubernetes scheduling approaches are not sophisticated enough to factor energy efficiency and infrastructure differences into the scheduling decisions. To address these shortages, we propose a novel scheduling system for Kubernetes clusters that considers infrastructure specifics and further task requirements. The system uses a benchmarking suite to model node performance, selects appropriate scheduling algorithms based on the number of tasks, and further optimizes cluster power consumption through re-scheduling. Our evaluation results show that the proposed approach achieves a power consumption reduction of18.46% compared to the Kubernetes scheduler.