Monitoring Kubernetes: Follow the Data [B]
summary
Datadog aggregates hundreds of trillions of anonymous metrics and shows concrete adoption and behavior numbers: 41% of Docker environments use Kubernetes as of Oct 2017, orchestrator use jumps above 50% once a site has 25 Docker hosts and 65% by 100 hosts, and 5% run multiple orchestrators. Kubernetes yields higher churn and density — container lifetime: 1.5 days on K8s vs 12 days on ECS and about nine application containers per host on K8s versus three on ECS — plus registry patterns and image pinning rates.
AI-written summary · about
chapters (31)
- 0:00Introduction
- 0:43About Datadog
- 1:38My role
- 2:26What we monitor
- 4:29Hacker News
- 5:19Containers are awesome
- 5:50Cloud Native Con
- 6:23Age of Kubernetes
- 6:53Linux
- 8:09Containerized Agent
- 8:45Datadog Data
- 9:02DockerHub Data
- 9:40Real Data
- 10:32Container Growth
- 11:18Orchestrators
- 11:37Docker Adoption
- 13:23New Data
- 13:54Adoption
- 15:02How does this impact us
- 15:31Host churn
- 16:23Dynamic containers
- 17:39Density efficiency
- 18:08Containers per host
- 18:32Application containers
- 19:22ECS agent
- 20:00Family Feud
- 20:35Nginx
- 21:38Datastores
- 22:04ECS vs Docker
- 24:54Metrics
- 25:20Resources
related talks
34mOptimizing operations & container management
20mKubernetes 101
37mYou’re Monitoring Kubernetes Wrong
43mBuilding Docker Images on Kubernetes: A Long Journey!
44mKubernetes at Datadog Scale
26mMonitoring in a Kubernetes World
05mWhy You Need Observability to Adopt Kubernetes at Scale
29mDocker for Managers