CAST-AI

CAST AI Kubernetes auto-optimization Spot ML.

Définition

CAST AI features : (1) Cluster Autoscaler : intelligent autoscaling EKS + AKS + GKE node groups based on pending pods + node utilization, optimal instance type selection mixed Spot + On-Demand + Reserved Instances, bin-packing automatic. (2) Spot Instances Optimization : up to 90% cost reduction via aggressive Spot adoption with automatic interruption handling, Spot diversification across instance types/zones reduces interruption risk. (3) Workload Autoscaler : automatically adjusts HorizontalPodAutoscaler + VerticalPodAutoscaler parameters based ML usage patterns. (4) Container Right Sizing : analyzes actual container CPU + memory usage vs requests, recommends right-sized requests, automatic applies via Mutating Webhook. (5) Commitment Manager : automated RI + Savings Plans recommendations + purchases via AWS Cost Optimization Hub integration. (6) Multi-cluster Cost Reporting + savings dashboard. Pricing : CAST AI takes percentage savings achieved (typically 5-25% of $ saved, no upfront cost), aligned with customer savings. Customers : Salesforce, Twitter, Akamai, T-Mobile, etc.

Origine

CAST AI fondee 2019 ; Series A $10M 2020 ; Series B $20M 2022 ; ~1500 customers 2024 ; ~$1B collective optimization 2024.

Exemple en contexte

Akamai Technologies Kubernetes Linode Akamai cloud platform deploye CAST AI sur 50+ clusters production, achieves 65% cost reduction via Spot Instances optimization automated + container right-sizing, ~$15M annual savings vs static configuration baseline.

Termes liés

• Kubecost / OpenCost — alternative monitoring.