Stop 05 · Storage ServicesTopics · 6/9

Data Archival (Glacier)

In short

Banks, hospitals and colleges must keep some records for years, even if nobody opens them. Archive storage keeps it for a fraction of normal prices, if you can wait to get it back.

Extremely low-cost, long-term storage for data you rarely need to access — retrieval can take minutes to hours depending on the tier, a trade-off worth making for compliance or historical data retention.

The blob character opens an old metal trunk full of bundled papers in a storeroom.

Think of it like this

Nani's trunk in the village

Old letters and certificates sit in a trunk at your nani's house. Storing them there costs nothing extra, but getting one means a phone call and waiting until someone visits. Archive storage is that trunk: cheap to keep, slow to fetch.

Watch it happen

Getting a file back from the archive

You

S3 Glacier

GET old-report.zip

1/5A normal GET on an archived object fails. The data isn't ready to be read directly.

The big ideas

What you need to know

Several archive tiers

Glacier Instant Retrieval reads in milliseconds. Flexible Retrieval takes minutes to hours. Deep Archive is cheapest and slowest.

Minimum storage time

Flexible Retrieval bills at least 90 days and Deep Archive at least 180, even if you delete early.

Retrieval costs money

Storing is cheap, but restoring lots of data, especially quickly, adds per-GB charges. Plan big restores.

Locks for compliance

Object Lock can make archives impossible to delete or change for a set period, which auditors like.

See it in code

Walk through the code

bash

1# Upload straight into Deep Archive2aws s3 cp results-2019.zip s3://college-archive/2019/ \3  --storage-class DEEP_ARCHIVE4 5# Later: ask for a 7-day readable copy6aws s3api restore-object --bucket college-archive \7  --key 2019/results-2019.zip \8  --restore-request '{"Days":7,"GlacierJobParameters":{"Tier":"Standard"}}'9 10# Check progress (look for ongoing-request="false")11aws s3api head-object --bucket college-archive --key 2019/results-2019.zip

1/3Pick the storage class at upload time, no lifecycle rule needed.

Restores run in the background. head-object shows a Restore field that flips to ongoing-request="false" when it's ready.

Save yourself the pain

Common mistakes

Oops

Archiving data you might need back within minutes in an emergency.

Do this instead

Match tier to RTO. If minutes matter, use Glacier Instant Retrieval or a warmer class.

Oops

Restoring a whole archive at the fastest tier 'just in case'.

Do this instead

Expedited and large restores cost more. Restore only what you need, with the slowest tier your deadline allows.

Your mission

Freeze and thaw a file

~20 min

Quick check

Why does a GET fail on an object in Glacier Flexible Retrieval?

Go deeper

1Move a sample object into an archival storage tier and note the retrieval options.
2Learn why archival storage fits compliance-driven retention (e.g. 7-year financial records).
3Understand the retrieval time trade-off before choosing archival storage for anything you might need urgently.