Getting the Optimal Service Efficiency That Autoscalers Won’t Give You
summary
An AI-driven, iterative optimizer applies Kubernetes and JVM settings, runs JMeter load tests, collects KPIs via Elastic APM, and uses reinforcement learning to propose new configs; after about experiment 34 / 90 hours it finds a configuration that increases CPU and memory requests, aligns JVM max heap inside container limits, prevents autoscaler churn, and achieves 49% cost reduction while keeping latency and SLOs within constraints.
AI-written summary · about
chapters (8)
- 0:00Introduction and challenges
- 2:24Kubernetes fundamentals
- 7:09Auto scaler evaluation
- 10:22AI-powered optimization
- 14:06Real-world case study
- 19:01Optimization results
- 21:52Resiliency and takeaways
- 25:53Q&A session
related talks
27mLeverage ML to automatically tune K8s apps
31mFailing to Autoscale?
31mAccelerate AI adoption: automated cloud efficiency
42mOptimize Kubernetes costs while matching your application SLO using AI
32mOptimizing Application Performance on Kubernetes
51mWhy You Need Performance Tests for Proper Kubernetes Scaling
22mIntelligent Autoscaling in Enterprise...
38mAutoscaling Containers... with Math