S3 CRR vs SRR cost: what changes (transfer, storage, requests)

If you're choosing between cross-region replication (CRR) and same-region replication (SRR), the cost difference usually comes down to one thing: whether replicated bytes are billed like cross-region transfer. This checklist helps you compare the real line items.

Start with the same 3 buckets

  • Replicated data (GB/month): the changed data copied from source to destination.
  • Replica storage: you store (almost) the same objects twice.
  • Extra requests: replication can increase PUT/GET/LIST/HEAD activity.

What changes between CRR and SRR

  • CRR: often includes cross-region transfer-like charges (or a per-GB replication fee, or both).
  • SRR: avoids cross-region transfer, but still duplicates storage and may add request costs.

Step 1: estimate replicated GB/month (changed data)

The biggest modeling mistake is using total bucket size. Replication volume is driven by writes and churn. Use this guide: estimate replication GB/month.

Step 2: price the per-GB replication line item

Use a per-GB assumption based on your plan:

Step 3: add replica storage

Treat replica storage as a separate ongoing line item (GB-month). Use Object storage cost calculator and add the replica stored GB as another storage line item.

Step 4: sanity-check request fees

For millions of small objects, request fees can matter. Use S3 request cost calculator or model requests in the Object storage cost calculator.

Validation checklist (before you choose CRR vs SRR)

  • Validate your goal: DR/multi-region reads (CRR) vs durability inside a region (SRR).
  • Validate replicated GB/month from write volume and churn (do not use bucket size).
  • Validate destination storage class and lifecycle policies (replica defaults matter).
  • Validate whether replication is billed as a feature fee, transfer, or both in your provider/region.

Worked comparison template (copy/paste)

  • Replicated GB/month = changed data per month (baseline + peak)
  • CRR per-GB fee = replication/transfer line item (if applicable)
  • SRR per-GB fee = replication line item (usually lower transfer component)
  • Replica storage = destination GB-month (by storage class)
  • Requests = extra PUT/GET/LIST/HEAD per month (many small objects scenario)

Common pitfalls

  • Comparing CRR vs SRR based on total bucket size instead of changed data per month.
  • Forgetting destination storage class (replica defaults to hot tier unless configured).
  • Ignoring peak replication windows (deployments/backfills can spike changed data).

Related guides

Sources


Related guides


Related calculators


FAQ

Is CRR always more expensive than SRR?
Usually yes because CRR often adds cross-region transfer-like charges, but the real driver is replicated GB/month (changed data). Storage and request fees can dominate depending on workload.
What is the main input for replication cost models?
Replicated GB/month (changed data), not total stored GB. Estimate it from write volume and churn, then price the per-GB replication/transfer line item.
Do I need to include destination storage?
Yes. Replication typically means storing a second copy of data. Destination storage is often the largest ongoing cost.

Last updated: 2026-01-27