SysPad › Lessons › storage lessons › S3 Glacier

How S3 Glacier works

Low-cost archival object storage. Instant Retrieval serves in ms; Flexible and Deep Archive trade retrieval latency (minutes to hours) for the lowest storage price. Models restore SLAs, not hot reads.

Some data you must keep for years but will almost never read, backups, compliance archives, raw logs. Paying hot-storage prices to let it sit is waste. Glacier is S3’s deep freezer: storage so cheap it’s almost free, in exchange for paying (and sometimes waiting) when you actually need it back.

What you will learn

What is archival storage optimised for, and what is the core trade?
What do the Glacier tiers differ on (especially retrieval time)?
How do retrieval speed options change cost and wait?
When is Glacier the right class, and when is it a trap?

Cheap to keep, costly to read

Glacier classes are part of S3, tuned for cold data: long-lived and rarely accessed. Storage costs a small fraction of S3 Standard. The catch is on the way out, you pay a retrieval fee, and for most tiers you wait for the data to be made available.

The Glacier tiers

Glacier Instant Retrieval: archival price with millisecond access (for cold data you occasionally need fast). Glacier Flexible Retrieval: cheaper, with retrievals in minutes to hours. Glacier Deep Archive: the cheapest of all, for data you’ll almost never touch, retrievals measured in hours.

⚓ Compliance and legal archives, long-term backups, media masters, raw data you keep "just in case."

Retrieval speed vs cost

For the non-instant tiers, retrieval offers speed options: Expedited (fastest, most expensive), Standard, and Bulk (slowest, cheapest). You pick how urgently you need the data versus how much you’ll pay to get it.

When Glacier helps, and when it bites

Glacier is ideal for data that’s big, long-lived, and seldom read. It’s the wrong choice for anything you read frequently or need instantly (outside Instant Retrieval), retrieval fees and latency will dwarf the storage savings.

Flashcards

What’s the core trade-off of Glacier vs S3 Standard?: Far cheaper storage in exchange for retrieval fees and (for most tiers) retrieval latency. Best for cold, rarely-read data.
Glacier Instant vs Flexible vs Deep Archive, the key difference?: Retrieval time/price: Instant = ms access; Flexible = minutes–hours; Deep Archive = cheapest, hours to retrieve.
Expedited vs Bulk retrieval?: Expedited = fastest, most expensive. Bulk = slowest, cheapest. You trade urgency against retrieval cost.
How do objects usually end up in Glacier?: Via an S3 lifecycle rule that transitions aging objects to a Glacier class automatically (e.g. after 90 days).
When is Glacier the wrong choice?: For frequently-read or latency-sensitive data (outside Instant Retrieval), retrieval cost/latency outweighs the storage savings.

Sources

Open the SysPad canvas · Official S3 Glacier documentation