Amazon EC2
Choosing a Purchasing Commitment: Savings Plans vs. Reserved Instances
- How likely the instance family/region is to change within the commitment term
- Current On-Demand spend baseline and its stability over the past quarter
- Whether Standard RI resale optionality matters as a hedge
- Existing overlapping commitments and their actual coverage (checked via Cost and Usage Report, not assumed)
Production premise
Reserved Instances commit to a specific instance family, size, region, OS, and tenancy in exchange for up to ~72% off On-Demand. Savings Plans commit to a dollar-per-hour spend instead of a specific configuration, trading some discount depth (Compute Savings Plans top out lower, though EC2 Instance Savings Plans match RI's ~72% within one family/region) for coverage that follows your usage instead of locking you to a resource shape.
Failure mode & inflection point
The commitment is real regardless of whether you use it — if your hourly Savings Plan commitment is higher than your actual eligible usage in a given hour, you're still charged the committed amount; unused RI capacity is dead spend unless it's a Standard RI you can resell on the Reserved Instance Marketplace (Convertible RIs can be exchanged internally; Savings Plans have no exchange or resale mechanism at all once purchased). Discount application order matters too: RIs apply first, then EC2 Instance Savings Plans, then Compute Savings Plans, with anything left over billed at On-Demand — teams that don't understand this order end up buying overlapping commitments that never actually stack the way they expected.
Production guardrail
Default to Compute Savings Plans for anything that isn't a genuinely static, unchanging fleet — the "at least this much useful discount, with room to change instance family/region later" trade-off beats a deeper discount you can't adjust when architecture or region needs shift. Reserve RIs (Standard, for resale optionality) for workloads you're confident won't move for the full term. Either way, this isn't a launch-day decision — put commitment coverage on a recurring review using Cost and Usage Reports, the same discipline Aurora's I/O-Optimized decision needs.