Cost a Glacier archive with per-object metadata overhead, retrieval fees and the early-deletion minimum duration.
Glacier classes add about 40 KB of index and metadata per object, so ten million small files carry roughly 380 GB of overhead you pay for on top of the data itself. Retrieval is billed twice, per GB restored and per restore request, and deleting before the 90-day (Flexible and Instant) or 180-day (Deep Archive) minimum still bills the remaining days. Prices here are user-supplied list prices — confirm them on the AWS S3 pricing page for your region. Archive tiers only pay off for large objects kept for a long time; the overhead and minimum-duration terms are what turn an apparently 10x cheaper tier into a more expensive one for small, short-lived files.
S3 Glacier Cost
billable GB = data + objects x 40 KB; cost = billable GB x rate + retrieved GB x retrieval price + retrieval requests ÷ 1000 x request price.
billable GB = data + objects x 40 KB; cost = billable GB x rate + retrieved GB x retrieval price + retrieval requests ÷ 1000 x request price. Glacier classes add about 40 KB of index and metadata per object, so ten million small files carry roughly 380 GB of overhead you pay for on top of the data itself. Retrieval is billed twice, per GB restored and per restore request, and deleting before the 90-day (Flexible and Instant) or 180-day (Deep Archive) minimum still bills the remaining days. Prices here are user-supplied list prices — confirm them on the AWS S3 pricing page for your region.
Archive tiers only pay off for large objects kept for a long time; the overhead and minimum-duration terms are what turn an apparently 10x cheaper tier into a more expensive one for small, short-lived files.
This calculator takes 8 inputs: Data archived, Objects archived, Archive storage price, Share of the archive retrieved each month, Retrieval price, Retrieval request price, Minimum billable storage duration, How long objects actually stay. The pre-filled defaults are a realistic starting point — replace them with figures from your own environment for a result you can act on.
Almost always. Packing many small files into a single archive object amortises the 40 KB per-object overhead and cuts retrieval request charges, at the cost of having to restore the whole container to read one member.