Deploying Kubernetes at scale
summary
Practical guidance: choose EKS as the default because it scales and offloads control plane ops, while Fargate obscures scaling and cost and DIY EC2 clusters make manageability your problem. Expect surprising constraints like namespace limits on huge nodes and hard tradeoffs when sizing container CPU and memory thresholds. On‑prem yields best performance on bare metal but VMs ease storage and SDN integration. Turbonomic claims agentless full‑stack visibility and will take prescriptive actions to resize VMs, pods, or EKS capacity. AWS runs ~84% of cloud Kubernetes.
AI-written summary · about
chapters (10)
- 0:00Introduction and setup
- 0:41Why build for scale now
- 3:13Architectural decisions on AWS
- 8:07Managed services comparison
- 9:30Container sizing and specs
- 10:31When and what to scale
- 11:35Full stack performance
- 12:18Q&A on-premise deployments
- 17:15Turbine AMA overview
- 19:41Closing and thanks
related talks
35mKubernetes on AWS: Zero to hero
17mKeynote: Kubernetes: Lessons from the Trenches
17mKeynote: Kubernetes: Lessons from the Trenches
33mKubernetes-Driven Multi-Cloud Deployments: Maximizing Scalability and Flexibility
29mThe Frugal k8s Engineer: Strategies for Amazon EKS
35mKubernetes on AWS: Practices & Opinions
16mKeynote: Cloud Native at AWS
24mGoing the Kubernetes Way: Self-hosted or a Managed Solution