Capacity Constraints Unveiled: Navigating Cloud Scaling Realities
summary
Elastic Cloud runs across 64 regions on four CSPs with roughly 40,000 hosts supporting about 50,000 deployments, and capacity shortfalls happen despite "limitless cloud" myths. The team turns stateful containers into a more stateless model, builds an autoscaling plane on Kubernetes using the cluster autoscaler, and relies on capacity reservations and GCP-style future reservations to guarantee nodes during peaks. They report lower maintenance impact after reserving fleet replacements and using instance-type priority fallbacks.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 1:30What is Elastic and Elastic Cloud
- 3:16Physical to cloud scaling
- 4:40Cloud promises and pitfalls
- 5:44Elastic Cloud’s scale challenges
- 8:16Stateful containers hurdles
- 10:24Why cloud constraints exist
- 12:11How we ended up with these cloud problems
- 16:10Redesigning Elastic Cloud
- 17:13Introducing Elasticsearch Serverless
- 18:38Kubernetes autoscaling
- 20:04Capacity reservations
- 21:47Planning with CSPs
- 23:16Proactive capacity strategies
- 24:49Take-home actions
- 26:27Availability zones and workloads
- 27:49Designing for growth
related talks
30mCapacity Management for Fun & Profit
20mManage Scaling, While Keeping Costs Under Control. Challenge Accepted.
14mKeynote: The Big Leap: Conquering the Legacy System and Building a Scalable Cloud Infra...
05mSponsored Keynote: The Cloud (And Your $$$) is Not Infinite: Dynamic Scheduling at Every Layer
43mCloud Native - effective computing at scale
31mKeeping Your Cloud Footprint in Check
21mDeploying Kubernetes at scale
27mFrom Supercomputing to Serving: A Case Study Delivering Cloud Native Foundation Models