Google Cloud Platform Podcast Borg and Kubernetes with John Wilkes
Oct 5, 2016
In this discussion, John Wilkes, Principal Engineer at Google and co-author of the Borg paper, dives into the fascinating evolution from Borg to Kubernetes. He explains how Borg's robust infrastructure supports large-scale systems, detailing its resilience and fault tolerance. John shares insights into the importance of open-source collaboration, the flexibility offered by Kubernetes for cloud deployment, and how it enables developers to focus on innovation. Plus, he touches on the excitement of new Google Cloud regions and even the tech behind Pokémon GO!
AI Snips
Chapters
Transcript
Episode notes
Borg's Dynamic Machine Allocation
- John Wilkes explains Borg allocates machines dynamically and manages failures automatically.
- Borg efficiently packs workloads, ensuring high machine utilization across Google's infrastructure.
Declarative Desired State Model
- Borg uses a declarative model where users state desired state and system reconciles actual state.
- This approach ensures resilience and automatic recovery from failures across the cluster.
Trust But Verify Systems
- Measure system behavior rigorously to verify fault tolerance and reliability.
- Monitor deviations and alert humans to intervene before issues escalate.