Intelligent Autoscaling in Enterprise...
summary
Company platform runs Kubernetes for 8,000 developers across 345 clusters with 1,000+ CPU cores and replaces ad hoc HPA tuning with an Intelligent Pod Autoscaler that emits actionable recommendations via a UI, accepts developer constraints (min/max and CPU/memory floors), integrates with Argo CD Jenkins and GitHub for rollout, and adds a timebased pod autoscaler CR that uses cron/absolute schedules to emit a target‑replicas metric (demo: scale 3 → 5 pods).
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:44Why we built autoscaling
- 3:32Kubernetes platform challenges
- 7:49Cost versus incidents trade-off
- 9:57User needs and customization
- 11:21Solutions by user type
- 13:08UI for autoscaling insights
- 15:15Time-based pod demo
- 19:22Open source lessons
- 20:24Conclusion and future
related talks
35mAutoscaling Kubernetes Deployments: A (Mostly) Practical Guide
31mFailing to Autoscale?
46mAutoscaling Pods and Nodes in Kubernetes: Where to Start?
17mScaling Made Simple: Implementing Autoscaling for Optimal K8s Performance
29mKubernetes Autoscaling: Beyond One Size Fits All
34mDeep Dive: SIG Autoscaling
35mSIG Autoscaling Intro
35mIntro: SIG Autoscaling