Meter
Retrieval makes cold data readable; egress moves it to you
Reading cold data and moving bytes to a recovery destination are separate steps. A restore may need both, plus API calls and temporary readable-copy storage.
Reviewed .
Put your own numbers through the calculator to see what this meter does to a normal month and to the day you restore.
Retrieval and egress measure different work
Two meters can apply to the same byte. Retrieval is the provider's charge for accessing data in an infrequent or archive tier, or for turning an offline object into a readable copy. Egress is the network charge for sending those readable bytes to the public internet, another region, or another destination class. [1] [3]
A typical backup restore from an archive to local hardware usually incurs both retrieval and internet egress. Restore and GET calls add request charges, and a temporary restored copy can add storage too. None of those read-only actions is an early deletion. [2] [1]
| Meter | What activates it | What it is not | Source ref |
|---|---|---|---|
| Retrieval | Reading an online cold object or restoring an offline archive, when the selected class publishes a retrieval rate [1] [3] | It is not a charge for crossing the provider's network boundary | [1] [3] |
| Egress | Moving readable bytes to a destination whose transfer rule is billable [1] | It is not the work required to make an archive readable | [1] |
| API requests | RestoreObject, GET, HEAD, LIST, multipart, or provider-equivalent calls [1] | Request count is separate from the number of bytes retrieved or transferred | [1] |
| Early deletion | Delete, overwrite, move, or transition before a published minimum, subject to provider-specific exceptions [1] [3] | does not apply to a read-only restore that leaves the source object in place | [1] [3] |
An archived restore crosses several billing checkpoints
An offline object first needs a restore request. Amazon S3 Glacier Flexible Retrieval and Deep Archive create a temporary readable copy; the archived source remains in its original class unless a separate copy or transition changes it. [2]
After readability is restored, a backup client issues GET or range requests and sends bytes to the recovery destination. Retrieval can therefore finish before egress starts, and a restore kept inside the same provider network can have retrieval without public-internet egress. [4] [1]
| Restore stage | Possible bill line | Operational test | Source ref |
|---|---|---|---|
| Request the archive | Restore request charge | How many objects or batches were submitted, and which speed was selected? [4] | [4] |
| Make data readable | Retrieval per GB and temporary-copy storage | Does the class charge retrieval, and how long is the readable copy retained? [1] [2] | [1] [2] |
| Read the objects | GET, HEAD, LIST, or equivalent request charges | How many objects, parts, metadata checks, and list pages did the tool touch? [1] | [1] |
| Move bytes to recovery | Egress or data transfer | Where did the bytes go, and which allowance and destination rule applies? [1] | [1] |
| Leave the source intact | Early deletion: does not apply | A temporary restore copy does not delete, overwrite, move, or transition the archived source [2] | [2] |
Sources
The explanations cite these sources. Each numbered reference in the text links to its entry here. Vendors change prices; if a date looks old, check the source. Three vendors publish a downloadable price feed and the rest are re-checked by hand, which is why the dates are not uniform.
Source pages checked4 sources
- 1Amazon Web Services: S3 pricingaws.amazon.com · checked 2026-07-26
- 2AWS documentation: Restoring objectsdocs.aws.amazon.com · checked 2026-07-26
- 3Google Cloud: Storage pricingcloud.google.com · checked 2026-07-26
- 4AWS documentation: Restoring objects retrieval optionsdocs.aws.amazon.com · checked 2026-07-26