proxy-KEDA2

True Scale-to-Zero with Intelligent Wake-Up Proxy
1-5 second cold starts • 100% memory reduction for idle services

1-5s

Cold start time

100%

Idle memory reduction

Vendor lock-in

Download CxO White Paper (PDF)

The Problem

Idle Kubernetes Costs

Modern Kubernetes platforms deploy hundreds of microservices, but most organizations run them always-on even when traffic is sparse or unpredictable, wasting 30-65% of compute budget on idle containers delivering no business value.

Current Solutions Fall Short

HPA keeps minimum replicas running. Serverless introduces lock-in and variable cold starts. Knative adds complexity. No production-ready pattern exists for true scale-to-zero with predictable wake-up.

Enterprise Reality

Average clusters run at 13-25% utilization. Internal tools and low-traffic services kept warm 24/7 represent $50K-$500K annual waste depending on cluster size and cloud pricing.

The proxy-KEDA2 Pattern

Kubernetes-native solution combining KEDA scale-to-zero, API-driven wake-up proxy, and activation catalogue. Zero vendor lock-in, sub-5s cold starts, seamless integration.

KEDA Scale-to-Zero

ScaledObjects drive true scale-to-zero using standard Kubernetes and KEDA. Idle workloads consume zero CPU/memory when not serving traffic.

API-Driven Wake-Up

Simple HTTP API (wake/app) patches KEDA and deployments, coordinates readiness, then hands control back to autoscaling. 1-5s from trigger to first request.

Activation Catalogue

Web portal lists applications with status. Users trigger wake-up then access service without kubectl or cluster access. ITSM/portal integration ready.

Kubernetes-Native

Runs on any CNCF-conformant cluster. No proprietary platforms, per-invocation billing, or serverless runtimes. Portable across on-prem, private cloud, managed K8s.

Business Value for CxOs & FinOps

Eliminate Idle Spend

100% memory/CPU reduction for dormant services. Turn always-on microservices into on-demand workloads aligned with actual business demand.

Production Performance

1-5 second cold starts keep within UX thresholds (0.1s instant, 1s uninterrupted, 10s acceptable). Full autoscaling preserved during active periods.

Zero Lock-In Risk

Standard Kubernetes + open-source KEDA. No proprietary serverless platforms or per-invocation pricing. Works with existing observability/ITSM investments.