Glacier vs Deep Archive cost: which S3 cold tier to choose

Quick answer

Amazon S3 has three archival tiers that trade storage price against how fast and how cheaply you can retrieve data. Glacier Instant Retrieval gives millisecond access at a higher storage price, for archives you occasionally need immediately. Glacier Flexible Retrieval is cheaper to store but retrieval takes minutes to hours. Glacier Deep Archive is the cheapest storage of all, around a dollar per TB per month range dynamics, but retrieval takes up to about 12 hours and costs the most per GB. Choose by how quickly you would ever need the data back, not just by storage price.

For data you must keep but rarely read, backups, compliance archives, raw logs, S3's Glacier tiers cut storage cost dramatically. But cold storage is a trade: the cheaper it is to store, the slower and pricier it is to get back. Choosing the right tier means being honest about how fast you would ever need the data, not just chasing the lowest storage number.

The three archival tiers

Tier	Storage price	Retrieval speed	Retrieval cost
Glacier Instant Retrieval	Higher of the three	Milliseconds	Moderate
Glacier Flexible Retrieval	Lower	Minutes to hours	Lower than Instant
Glacier Deep Archive	Lowest (about a dollar per TB per month)	Up to about 12 hours	Highest per GB

Glacier Instant Retrieval keeps archives you might need immediately, millisecond access like Standard-IA but cheaper to store, at a higher storage price than the other two Glacier tiers. Glacier Flexible Retrieval is cheaper to store with retrievals ranging from a few minutes (expedited) to several hours (bulk). Glacier Deep Archive is the cheapest object storage AWS offers, in the range of roughly a dollar per TB per month, but retrievals take up to about 12 hours and cost the most per GB.

Retrieval cost and time are the tradeoff

The mistake is optimizing storage price alone. Deep Archive is dramatically cheaper to hold than Standard, but if you retrieve frequently, the per-GB retrieval fees and request costs can erase the savings, and the multi-hour latency may be unacceptable. Ask two questions per data set: how often will I read it, and how fast do I need it back. Rare and patient means Deep Archive; rare but occasionally urgent means Instant Retrieval. This is the same cold-storage logic as GCP storage classes.

Minimum durations and per-object overhead

The Glacier tiers impose minimum storage durations (Deep Archive around 180 days, Flexible and Instant around 90 days), so deleting early still bills the minimum. There is also a small per-object storage overhead, which makes archiving millions of tiny objects less efficient than fewer larger ones. Aggregating small files before archiving avoids paying overhead per object, a cousin of the snapshot cost creep problem where small artifacts quietly accumulate.

Using lifecycle policies to tier automatically

The practical way to use these tiers is S3 Lifecycle policies that transition objects from Standard to Standard-IA to a Glacier tier as they age, and expire them when retention ends. That way hot data stays fast and cold data drifts into cheap storage automatically. Decide the target tier by the data's access pattern and required retrieval speed, mind the minimum durations, and check whether it is truly cheaper than starting in a warmer tier like the ones covered in S3 storage basics. Price your storage footprint against the resource catalog before committing large archives.

FAQ

What is the difference between Glacier and Deep Archive?

They are S3 archival tiers trading storage price against retrieval speed and cost. Glacier Instant Retrieval gives millisecond access at a higher storage price; Glacier Flexible Retrieval is cheaper to store with retrievals taking minutes to hours; Glacier Deep Archive is the cheapest storage (around a dollar per TB per month range) but retrievals take up to about 12 hours and cost the most per GB.

How cheap is S3 Glacier Deep Archive?

Deep Archive is the cheapest object storage AWS offers, in the range of roughly a dollar per TB per month, dramatically less than S3 Standard. The tradeoff is retrieval: getting data back takes up to about 12 hours and carries the highest per-GB retrieval fees of the Glacier tiers, so it is only economical for data you almost never need and can wait for.

When should I use Glacier Instant Retrieval instead of Deep Archive?

Use Instant Retrieval when you rarely access the data but need it back immediately when you do, since it offers millisecond access like a warm tier at a lower storage price. Use Deep Archive when you can tolerate multi-hour retrieval latency and want the absolute lowest storage cost. The choice hinges on required retrieval speed, not just storage price.

Do Glacier tiers have minimum storage durations?

Yes. Deep Archive bills a minimum of around 180 days and the other Glacier tiers around 90 days, so deleting or moving an object earlier still incurs the minimum-duration charge. There is also a small per-object overhead, which makes archiving many tiny objects less efficient than aggregating them into fewer larger objects before archiving.

Can retrieval fees erase Glacier savings?

Yes. If you retrieve archived data frequently, the per-GB retrieval fees and request costs, highest on Deep Archive, can exceed the storage savings you gained by moving to a cold tier. Cold storage is cheap to hold and expensive to touch, so it only saves money for data you genuinely read rarely. Model retrieval frequency, not just storage volume.

Does C3X help with S3 storage tiering costs?

C3X reads your infrastructure code and prices the storage resources you define against a live catalog before deploy, so you can compare the monthly cost of S3 storage classes and Glacier tiers for a data set in the pull request. That helps you choose a tier by cost before committing large archives, factoring in the storage price differences up front.

What to do next

Choose the right S3 cold tier by cost before you archive. C3X prices your storage from code against a live catalog. Start with the quickstart and the resource catalog.