Cloud egress: S3 Transfer Acceleration vs Internet DT-out
S3 Transfer Acceleration is a distinct path class from standard Internet DT-out: acceleration uses CloudFront edge networking with its own fee schedule, while ordinary bucket downloads meter Internet data-transfer-out. Model them separately — Unknown until verified.
Updated
Transfer Acceleration is a distinct path class from standard Internet downloads — model acceleration fees separately from DT-out.
Path matrix: S3 Transfer Acceleration vs Internet DT-out (educational)
Dated starting points only — Unknown until verified means no invented $/GB.
Teams enable Transfer Acceleration for distant uploads/downloads and then paste standard Internet egress rates into the same cell. PipeToll keeps acceleration as its own path template.
Path templates to model
Global mobile upload into a regional bucket — acceleration may help latency; fee is not “free CDN.”
Public asset GET without CloudFront — standard Internet DT-out; cliff math applies.
CloudFront in front of S3 — different from Transfer Acceleration; use origin-fill literacy.
Same-region ETL pull — usually not an acceleration candidate; wrong rate class.
Mixed clients — some accelerated, some not — split bytes before Mode A.
Common traps
Assuming acceleration replaces CloudFront. Blending accelerated GB with standard DT-out. Ignoring that acceleration does not erase request ops. Treating latency wins as automatic cost wins — no savings guarantee.
FinOps checklist
Inventory which prefixes/clients use accelerate endpoints.
Split standard Internet vs accelerated bytes for 30–90 days.
Price each class with dated S3 cites; badge Unknown otherwise.
