How Ticketmaster runs Kubernetes for 80% less without managing VMs
summary
Ticketmaster moves its Kubernetes worker fleet onto Spotinst Ocean to run 100 percent of EC2 on Spot Instances with programmatic fallback and eliminate manual VM ops. Results: a 5x improvement in resource utilization, about 60% cost reduction versus prior VM setups, and cluster packing that reaches ~80% CPU and memory allocation. The team automates container-driven node sizing, heterogeneous instance families, fast node boot (3–4 minutes), and safe fallbacks to On‑Demand when spot capacity vanishes.
AI-written summary · about
chapters (14)
- 0:00Intro
- 0:30About Aerco
- 4:15Kubernetes Anatomy
- 7:30Instant Sizing Layer
- 9:58Container to Even Scaling
- 14:38Instance Pricing
- 17:39Cost Savings
- 19:52Container Orchestration
- 26:53DevOps
- 32:33Deployment
- 35:58Application
- 41:28Lessons learned
- 44:11How things change over time
- 51:11Summary
related talks
34mBringing 1976 into 2016: Ticketmaster’s Public Cloud and Kubernetes Strategy
34mCloud Native Transformation: How Ticketmaster is Changing Culture and Speeding Up
44mVax to K8s: Ticketmaster's Transformation to Cloud Native Devops
34mKubernetes: Reducing Infrastructure cost and complexity
14mKeynote: Lessons and Rhythms from the Hybrid Cloud
39mKubernetes – Infinite Space for Your Tests
34mReduce Costs and Optimize on Your Cloud with Karpenter
39mOptimizing Resource Usage in Kubernetes