Work out provisioned vs on demand capacity instantly with clear inputs, formula shown and shareable results.
Provisioned capacity must be sized for peak but is billed constantly, so its economics depend entirely on the ratio of average to peak. On-demand costs more per request but nothing when idle, which makes it cheaper for spiky workloads below roughly 15 to 20 percent average utilisation of peak — and far more expensive for steady high-volume ones.
Mode comparison
provisioned = peak units x unit rate x 730; on-demand = average requests x per-request rate; utilisation = average / peak
Partly. It tracks demand with a lag of minutes, so sudden spikes are throttled. On-demand absorbs instant bursts up to double the previous peak.
Yes, though usually only once every 24 hours. Switching to on-demand ahead of an unpredictable event and back afterwards is a common tactic.