Cloud Native - effective computing at scale
summary
Containers cut VM overhead so Google runs about 4 billion containers per week and orchestration moves scheduling, autoscaling and image delivery off operators to systems like Borg → Kubernetes (clusters up to ~5,000 nodes). Kubernetes adds pods, labels/selectors and extensibility; Istio injects Envoy sidecars for canary routing, mTLS and telemetry; autoscalers and cluster autoscaling handle spikes. Options for durable state include Cloud Spanner, Cassandra or Vitess, and gVisor adds a VM-like kernel boundary when stronger isolation is needed.
AI-written summary · about
chapters (18)
- 0:00Start
- 1:43Virtual Machines
- 5:24Control Groups Unix Kernel
- 7:39Fast Deployment
- 13:37Requirements
- 19:07Separation of Development and Operations
- 19:47Case Study
- 21:14What Is Kubernetes
- 21:42Kubernetes
- 23:01Persistent Storage of Kubernetes
- 23:41Decomposed the Kubernetes Master
- 25:07Labels and Selectors
- 30:06Rollout of the New Features of the Steady State
- 31:59Hybrid Cloud
- 32:32Cloud Native Computing Foundation
- 36:53Is Kubernetes and Overkill for Small Projects
- 37:55Networking in Kubernetes
- 39:28G Visor
related talks
23mMaking Cloud Native Computing Universal and Sustainable
31mReverse Engineering Cloud Native: Interoperability and Community
42mCloud Native Applications
27mCloud Native: why, oh why?
07mKeynote: Exploring the Frontiers of Cloud Native
08mKeynote: Cloud Native in the Enterprise: Production, Best Practice and Innovation
47mCloud Native and Sustainable, Reproducible Scientific Computing
23mKeynote: A Brief History of the Cloud from Servers to VMs to Buildpacks to Cloud Native Containers