S3 Cross-Region Replication (CRR) is a distinct path class from same-region Internet GETs and Transfer Acceleration: CRR meters inter-region replication bytes (and ops) from source to destination bucket, while public downloads meter Internet DT-out and TA uses its own acceleration fees. Model them separately — Unknown until verified.
Updated
Cross-region replication is not end-user egress — keep CRR, same-region GET, and Transfer Acceleration on separate lines.
Path matrix: S3 CRR vs same-region GET vs Transfer Acceleration (educational)
Dated starting points only — Unknown until verified means no invented $/GB. Region-pair CRR rates vary; always cite the destination Region on the S3 pricing page before Mode A arithmetic.
Teams enable CRR for durability or multi-region reads, then paste Internet egress rates into the replication cell — or blend CRR with Transfer Acceleration. PipeToll keeps CRR as its own path template.
Path templates to model
Primary → DR bucket CRR — inter-region replication bytes; ops and storage still separate.
Public asset GET (same region) — standard Internet DT-out; cliff math applies.
Global upload via Transfer Acceleration — acceleration fees, not CRR.
CloudFront in front of S3 — origin-fill literacy; different from CRR.
Same-region ETL pull — usually not CRR; wrong rate class if you paste inter-region $/GB.
Common traps
CRR called “egress to users” — Calling CRR “egress to users” mixes replication meters with Internet GET. Users download on a different path template.
Replication blended with Internet GET — Blending replication GB with Internet GET GB on one chart hides two classes. Run Mode A twice — ?path=s3-crr vs Internet DT-out.
TA as free CRR substitute — Treating Transfer Acceleration as a free CRR substitute invents architecture. TA is an upload/download acceleration fee, not multi-region durability — see the TA spoke.
Ops lines ignored — Ignoring replication PUT / LIST / ops lines understates the CRR path. Bytes and request meters stay separate.
SRR ≠ CRR — Same-Region Replication stays inside one Region; Cross-Region Replication crosses Regions and bills inter-region replication meters. Never paste SRR into a CRR cell.
Transfer Acceleration fee cell — Unknown; see the TA spoke.
Replication / request ops remain on their own lines. Run three Mode A estimates. No savings guarantee.
Numbered S3 CRR vs GET vs TA stack — Unknown rate cells
#
Path class
Volume
Unit rate
1
S3 Cross-Region Replication
1,200 GB
Unknown $/GB
2
Same-region Internet GET
4,000 GB
Unknown $/GB
3
Transfer Acceleration
300 GB
Unknown $/GB
4
S3 request / replication ops
as measured
Unknown
CE / CUR mapping: Usage types that look like “Data Transfer” + S3 replication / inter-region toward another Region belong on this spoke — not only the generic cross-AZ vs cross-region matrix. Deep-link ?path=s3-crr so Mode A labels the CRR path context.
Second educational stack: 450 GB CRR + 80 GB same-region compute GET (no public Internet)
#
Path class
Volume
Unit rate
1
S3 Cross-Region Replication
450 GB
Unknown $/GB (region pair)
2
Same-region compute GET
80 GB
Unknown — usually not Internet DT-out
3
Replication PUT / LIST ops
as measured
Unknown
Run Mode A twice (CRR vs compute GET). Do not paste Internet $/GB into the CRR cell. No savings guarantee.
SRR ≠ CRR worked clarity: Same-Region Replication stays inside one Region (no inter-region replication DT class). Cross-Region Replication crosses Regions and bills inter-region replication meters plus ops. In Mode A, run ?path=s3-crr only for CRR bytes — never blend SRR, Internet GET, or Transfer Acceleration into that cell.
Hold the 1,200 GB CRR + 4,000 GB same-region GET baseline and walk three replication / delivery runs. CRR, Internet DT-out, Transfer Acceleration, and ops $/GB cells stay Unknown until dated S3 cites — literacy arithmetic, not a quote.
Run A — CRR + Internet GET split: 1,200 GB CRR inter-region + 4,000 GB same-region Internet GET. Open Mode A (?path=s3-crr&gb=1200) for CRR; second Mode A for GET. Never one blended $/GB.
Run B — CRR-only twin (no public GET): same 1,200 GB CRR with compute/same-region reads instead of public Internet GET — Internet DT-out → ~0 on that half. Re-run Mode A only for measured CRR GB; keep ops on their own line.
No. CRR replicates objects between buckets across regions with inter-region replication meters. Public same-region GETs meter Internet data-transfer-out. Keep them on separate lines.
Is Cross-Region Replication the same as Transfer Acceleration?
No. Transfer Acceleration optimizes client↔S3 transfers via edge networking with its own fees. CRR is bucket→bucket replication across regions — a different path class.
Does enabling CRR guarantee lower bills?
No. PipeToll never guarantees savings. CRR adds replication meters; resilience and cost are different questions. Educational literacy only.
How do Cost Explorer / CUR labels map to CRR?
Look for S3 Cross-Region Replication / inter-region replication talkers — not only “Regional Data Transfer.” Map them on the Cost Explorer decoder, then open this spoke and Mode A (?path=s3-crr). Keep CRR separate from InterZone and Internet DT-out.
Is Same-Region Replication (SRR) the same as Cross-Region Replication (CRR)?
No. SRR copies objects between buckets in the same Region (different meters / usually no inter-region DT). CRR copies across Regions and meters inter-region replication. Do not paste Internet DT-out or SRR assumptions into the CRR cell — see glossary: S3 CRR.
Trademarks: Amazon Web Services, AWS, Microsoft Azure, Google Cloud Platform, GCP, and Cloudflare are trademarks of their respective owners. PipeToll (Ameti One LLC) is not affiliated with, endorsed by, or sponsored by those companies. Educational comparisons may be outdated — verify current rates with each provider.
