Skip to content

ECO-AIM-AI-005

Name: Always-on inference endpoints

Category: AIM

Family: AI

Primary layer: ai

System layers: ai

Description

Always-on endpoints waste baseline compute when idle.

Impact

  • confidence: 0.6
  • notes: High for GPU-backed endpoints.
  • type: carbon

Detection

  • languages:
  • infra
  • method: trace

Remediation

  • guidance: Autoscale; scale-to-zero where feasible; consolidate endpoints.
  • tradeoffs: Cold starts and scheduling complexity.

Pattern examples

No pattern examples provided.

Remediation examples

No remediation examples provided.