Papers on Cluster Management at Scale
Published:
Insightful Papers on Cluster Management at Scale
The papers presented below offer a historical perspective on the creation and expansion of Google’s internal system “Borg”. Several of the techniques and procedures that were researched and implemented within “Borg” have since been adopted into Kubernetes. What these papers particularly highlight are the important factors and trade offs that must be taken into account when managing a global cluster. The management of such a cluster is a substantial undertaking, and even small improvements in utilization can yield savings amounting to millions of dollars.
- Autopilot: workload autoscaling at Google
- Borg: the Next Generation
- Large-scale cluster management at Google with Borg
- CPU performance isolation for shared compute clusters