Reserved Instances vs Savings Plans: FinOps decision for PR
When to commit to RI vs Savings Plans based on a growing SMB stable workload.
Stable AWS bill three consecutive quarters — signal to evaluate commitment discounts. Reserved Instances (RI) and Savings Plans (SP) reduce compute 30-72% in exchange for 1-3 year commitment. Growing Puerto Rico SMBs fear lock-in; mature FinOps models scenarios before signing. AWS publishes guidance in FinOps blog and billing documentation — read before impulse RI purchase after high bill.
Flexible Compute Savings Plans across instance family and region — better for heterogeneous workloads (mix EC2, Fargate, partial Lambda discount). Standard instance-specific RI maximizes discount if predictable workload — same m5.xlarge 24/7 in prod. Convertible RI if anticipating instance type migration in 12 months.
Set a coverage target for the baseline
Coverage target: 70-80% stable baseline, not 100% — keep On-Demand for burst and experimentation. Cost Explorer coverage report shows gap; over-commitment pays for unused capacity if post-contract downsizing. Mainland companies with us-east-1 DR for PR operations should aggregate cross-account usage before organization-wide SP.
Do not compromise security for savings: downsizing instance until PHI app performance degrades is not FinOps, it is risk. Bruce Schneier does not write FinOps, but informed trade-offs principle applies — document RI vs SP decision in committee with finance and IT.
Review your commitments every quarter
Review quarterly: AWS Compute Optimizer + CUR data export. San Juan SMB that doubled users post-pandemic may have outgrown 2022 RI — marketplace sell unused RI or convert. FinOps decision is continuous process, not annual checkbox.
Operations and execution — connecting strategy with what the team ships every week.
