DynamoDB Reserved Capacity: Buying Strategy, Coverage, and Ladder Approach

DynamoDB Reserved Capacity is structurally similar to Reserved Instances for EC2 but receives far less attention because the cost surface looks small until you do the math. At enterprise scale (above $100K/month DynamoDB spend), a disciplined reserved capacity strategy routinely cuts the bill by 40 to 55 percent. The buying mechanics are simple; the strategy is where the leverage hides.

What Reserved Capacity covers

DynamoDB Reserved Capacity is a commitment to a baseline WCU/RCU quantity for 1 or 3 years, in exchange for a discount against the standard provisioned rate. The commitment applies per region and is shared across all tables and GSIs in that region.

Eligibility:

• Provisioned capacity only. On-demand consumption is not eligible.
• WCU and RCU are purchased separately.
• Applies to base tables and GSIs equally.
• Reserved capacity is regional; no cross-region sharing.
• Replicated capacity in Global Tables is not reserved-eligible.

Term and payment options

Unlike EC2 RIs, DynamoDB Reserved Capacity has no no-upfront or partial-upfront option. The commitment is paid in full at purchase.

The buying strategy

The four-step process we run with every client entering DynamoDB reserved capacity:

Step 1: baseline measurement

Pull 90-day CloudWatch data for ProvisionedReadCapacityUnits and ProvisionedWriteCapacityUnits across all tables and GSIs in each region. Identify the 25th-percentile baseline: the floor below which provisioned capacity rarely drops.

Step 2: coverage target

Target 60 to 80 percent reserved coverage of the 25th-percentile baseline. The remainder absorbs short-term traffic shifts and auto-scaling movement. Going above 90 percent risks paying for capacity that gets retired.

Step 3: ladder versus one-shot

For 3-year commitments, ladder purchases across 6 to 12 months rather than buying the full commitment in one transaction. The ladder protects against:

• Estate retirement before the 3-year term ends.
• Capacity-need miscalculation.
• AWS price changes during the buying window (rare but it happens).

Step 4: WCU and RCU treated separately

Workloads are rarely balanced between reads and writes. Buy WCU and RCU coverage independently against measured baseline for each. Buying matched quantities is a common error that over-buys one and under-buys the other.

Coverage management across the term

Reserved capacity coverage is not set-and-forget. Quarterly review:

• New tables added since last review; does the baseline need increasing?
• Tables migrated to on-demand; does coverage need reducing?
• GSI changes affecting baseline capacity?
• Regional shifts that change which region the baseline lives in?

The discipline keeps coverage aligned to actual baseline as the estate evolves.

The on-demand to provisioned conversion

Reserved capacity is most powerful as the second move after converting eligible on-demand tables to provisioned. The combined sequence:

1. Identify on-demand tables with sustained utilisation above 18 percent of peak.
2. Convert to provisioned with auto-scaling. Immediate ~7x saving on the converted portion.
3. Wait 30 days for new provisioned baseline to stabilise.
4. Purchase 1-year or 3-year reserved capacity covering 60 to 80 percent of the new baseline. Additional 50 to 80 percent saving on the covered portion.

Buyers who skip step 4 leave most of the available savings on the table.

GSI capacity considerations

Reserved capacity covers GSI throughput identically to base table throughput in the same region. Two operational notes:

• GSI capacity is part of the baseline calculation; do not exclude it.
• If you delete a GSI mid-term, your reserved coverage remains in force; the now-unused capacity stays committed until term end. Audit reserved capacity before deleting large GSIs.

Exit risk and mitigation

Reserved capacity is non-refundable and non-transferable across regions. The exit risks:

• Regional migration. Moving a workload from us-east-1 to eu-west-1 mid-term leaves stranded us-east-1 reserved capacity.
• Table retirement. Decommissioning a major table leaves over-provisioned reserved capacity.
• Capacity-mode shift. Switching a table back to on-demand strands reserved capacity.

Mitigations: ladder purchases, 1-year commitments for newer or less-stable workloads, 3-year only on workloads with confident 3-year footprint.

The EDP negotiation layer

Reserved capacity discounts are published; the EDP layer adds:

• Custom reserved capacity rates. For sustained DynamoDB spend above $100K/month, AWS will discount the published reserved capacity rates by an additional 10 to 25 percent.
• Custom on-demand pricing. For workloads that cannot move to provisioned, custom on-demand rates 15 to 35 percent below standard.
• Backup and Global Tables transfer discounts.

Worked example: $720K DynamoDB estate

A logistics customer with $720K annualised DynamoDB spend, all on-demand. Audit:

• Average utilisation 42 percent of peak across the estate.
• 14 tables, 2 regions, 31 GSIs.
• Zero reserved capacity coverage.

Sequence:

1. Converted 11 of 14 tables to provisioned with auto-scaling. Annual cost dropped from $720K to $148K (a 79 percent reduction from the on-demand-to-provisioned shift alone).
2. Measured new baseline over 60 days.
3. Purchased 3-year reserved capacity at 65 percent coverage. Additional $86K annualised saving.
4. Negotiated 18 percent custom discount on remaining on-demand at EDP renewal. $11K saving on residual on-demand spend.

Final annualised cost: $51K, a 93 percent reduction from baseline. The reserved capacity layer alone saved over 50 percent of the post-conversion spend.

Action checklist

1. Pull 90-day CloudWatch data for provisioned WCU and RCU per table per region.
2. Identify 25th-percentile baseline. This is your reserved-capacity target.
3. Convert sustained-utilisation on-demand tables to provisioned first.
4. Buy WCU and RCU reserved capacity separately at 60 to 80 percent of baseline.
5. Ladder 3-year purchases across 6 to 12 months.
6. For sustained DynamoDB spend above $100K/month, scope custom reserved-rate discount into EDP renewal.
7. Quarterly reserved coverage review against current baseline.
8. Contact our advisory team for a DynamoDB reserved capacity buying plan benchmarked against $2.4B+ of reviewed AWS spend.

DynamoDB Reserved Capacity is the most cost-effective AWS database commitment by ratio. See our DynamoDB pricing strategy guide for the broader DynamoDB cost picture and the database cost strategy guide for the full database stack context.