Practical Zombie Hunting for Kubernetes Users
Talkat KubeCon + CloudNativeCon Europe 2025 · on Thursday, 2025-04-03· at 16:00· Level 1 | Hall Entrance N10 | Room F·
summary
Claims that ~25–30% of servers do no useful work and average utilization sits at ~12–18%, costing an estimated $26B in always‑on cloud waste in 2021 and examples like Twitter’s forgotten 700 GPUs and personal clusters costing £1,000/month. Detection plus safe removal is key: use LightSwitchOps (infrastructure as code, idempotent restore), FinOps/Backstage cost insights, automated leases and simple scripts, or tools like DailyClean and Kruize Autotune. Try the eco-monkey scream test, aim for 70–80% utilization, and expect 12–50% savings from basic automation.
AI-written summary · about
chapters (6)
- 0:00Introduction
- 3:12Zombie server problem
- 8:56Zombie servers in Kubernetes
- 15:28Detecting zombie resources
- 24:27Sustainable server ops
- 29:43Call to action
related talks
47mCost optimization
33mUncover Kubernetes cost-optimization trends
33mKubernetes cost optimization - How to actually do it
14mCut Costs, Not Performance: Implement Kubernetes Cost Optimization Strategies
41mFinOps and Cloud Cost Optimization
05mFive Minute Cost-Optimisation
56mA Developer's Perspective On Greener Future: GreenOps in practice
43mKubernetes cost optimisation - How to actually do it