How an AI driven approach reduces cloud cost and makes your Kubernetes infrastructure autonomous
summary
An AI-driven engine combines KEDA horizontal scaling, VPA recommendations and a cluster autoscaler that picks spot instances by CPU/memory requirement to cut waste; a customer case finds about $1.2M annual savings and spots yield ~40–80% discounts. A live demo processes 150 files, the system launches c5a 2x and 4x nodes and adapts while Amazon takes ~2 minutes to provision new VMs. Recommendation: prefer dynamic spot selection over reserved instances or savings plans.
AI-written summary · about
chapters (28)
- 0:00Intro
- 0:50The problem
- 2:40Cloud providers
- 4:00Clock Computers
- 4:38Reserve
- 5:39Why Reserve
- 6:51IDLE capacity
- 7:39Manage Interrupt
- 8:41AWS example
- 13:06The 3 questions
- 15:48Horizontal photoscaling
- 17:55Vertical autoscaling
- 19:29DPI example
- 20:08HP vs VPN
- 21:05Cluster Autoscaler
- 22:17Before and After
- 23:02What the AI engine recommend
- 24:06Cost optimization
- 25:32Demo
- 25:42App description
- 26:37Small batch process
- 26:52Amazon SQL
- 28:41First VM
- 32:27CPU utilization
- 32:37CPU supply
- 33:33CPU supply decrease
- 34:17Automation with HPA
- 36:55Conclusion
related talks
31mAccelerate AI adoption: automated cloud efficiency
33mUncover Kubernetes cost-optimization trends
21mBest Practices for Optimizing Your Cloud Spend with AI
26mDescubra las tendencias de optimización de costos en Kubernetes
33mThe Big Kubernetes Mistake Paying More Than Necessary
31mFailing to Autoscale?
12mAI-Driven Platform Engineering
22mAI in Kubernetes The Promise