arXiv.itsnick.co

an independent field guide to 23 million papers from arXiv & OpenAlex — not affiliated with arXiv.org

Computer Science · Computer Networks and Communications

Dynamic Kubernetes Scheduling: Load and Resource Optimization in Collaborative Edge Computing

Tong Liu, Fanping Zeng, Pengcheng Xia

See on the map

Abstract

Kubernetes, as a mainstream container orchestration platform, is widely used to manage microservice-based applications in cloud environments. However, native scheduler is relatively single in resource scheduling, and only considers CPU and memory, which cannot meet the needs of collaborative edge computing for refined resource scheduling. In addition, it only relies on the resource requests of containers for scheduling, and does not fully consider the actual node load, which may lead to load imbalance problem. To solve these problems, a flexible and dynamic task scheduler, FlexiTask scheduler, is designed. Based on the real-time load and historical resource utilization data of nodes, the scheduler constructs a multi-index model that adapts to the task requirements of collaborative edge computing. Its core goal is not only to achieve cluster load balancing, but also to improve system stability and overall resource utilization efficiency in the environment of node load fluctuation and resource change. The experimental results show that compared with the native scheduler and Tencent Crane scheduler, FlexiTask scheduler improves the task deployment efficiency by 30.5% and 11.2% respectively, reduces the cluster resource imbalance by 44.1% and 34.0%, and in specific application scenarios, the task scale is increased by 96.0% and 22.5%, which fully demonstrates its excellent performance.

More like this