On-Demand vs Provisioned Capacity

~8 min read

Pay-per-request versus committing to a baseline throughput, and how Auto Scaling bridges the gap.

On-Demand vs Provisioned Capacity is a Pro topic

Sign in, then upgrade to Pro or Power to unlock this topic and the full AWS curriculum.

Key points

  • On-Demand: pay per request, zero capacity planning, higher per-request cost, best for new/unpredictable workloads
  • Provisioned: commit to RCU/WCU throughput, cheaper if well-utilized, requests beyond capacity are throttled without Auto Scaling
  • Auto Scaling adjusts Provisioned capacity automatically within configured bounds based on utilization
  • Switching between modes is limited to once per 24 hours — plan ahead rather than reactively