StormForge

StormForge applies ML rightsizing patches directly to live Kubernetes deployments.

StormForge analyzes 14 days of utilization metrics with ML models trained on historical patterns, predicts the right CPU and memory requests and limits, and either patches resources via the Kubernetes API or exports YAML for GitOps. Node Optimization extends the same predictive approach to cluster autoscaler decisions, and JVM heap size recommendations cover Java workloads explicitly.

Who it's for. Platform teams of 5 to 30 engineers running Kubernetes with chronically over-provisioned requests where 30 to 50 percent of cluster capacity is wasted. Scenario: a 100-node cluster sits at 15 percent average CPU utilization because pods request 4 cores and use 0.5, StormForge analyzes the metrics, applies rightsizing automatically, and the cluster runs on 30 fewer nodes.

Tradeoffs. Paid only with no free tier; PAYG exists but is sales-only. Kubernetes only; no VM rightsizing, no database tuning beyond JVM heap. Automated patching of live deployments requires staging validation and trust. Node Optimization overlaps with CAST AI but lacks the spot orchestration story.

Compare: CAST AI, Karpenter, Zesty, Densify