Comparison
On-demand pricingvsReserved capacity
On-demand pricing
you pay the full hourly rate for capacity you can give back at any second, which is the most expensive way to run anything steady.
Paying the published rate per unit of time with no commitment and no interruption risk. It is the right price for genuinely unpredictable or short-lived workloads and the wrong one for a baseline that has run continuously for two years. Most cloud overspend is simply a steady-state fleet still being billed this way because nobody did the commitment exercise.
Full entry →Reserved capacity
you promise a year of spend in advance and pay perhaps forty percent less, and you owe it whether or not you use it.
Committing to a level of usage over one or three years in exchange for a substantial discount. It is close to free money for a stable baseline and a liability if the architecture changes underneath it — a migration that halves your compute leaves the commitment intact. The practical approach is to commit only to the floor you are confident of and leave the variable part on demand.
Full entry →