Amazon Simple Storage Service Glacier
Low-cost long-term archive
S3's archive storage classes (Glacier Instant Retrieval, Flexible Retrieval, Deep Archive) are very cheap for rarely accessed data; in exchange there are retrieval fees and retrieval times from milliseconds to many hours. The old vault-style Glacier service (separate API) no longer accepts new customers; use the S3 storage classes instead.
- Cost$ Low
- Difficulty●●●●●
- Costs money when idle?A little
- Serverless
- In examsCLF · SAA · SOA
Think of it asLike putting things in a storage facility outside the city: storage is cheap, but getting them back means waiting and paying for delivery.
When to use it
✓ Use it when
Storing data for many years that is rarely read: backups, compliance logs, archival data.
✗ Avoid it when
Frequently read or short-term data → S3 Standard / S3 Intelligent-Tiering.
Common uses
- Keeping logs and records for years per regulations
- Replacing backup tapes
- Archiving original media
What does it cost?
- Pricing model: Charged per GB-month stored, number of requests and GB retrieved (depending on retrieval speed).
There is a minimum storage duration (early deletion is still charged in full) and a per-object overhead fee, so many small files get very expensive.
When the question says long-term, rarely read, cheapest, and waiting hours is fine, choose Glacier Deep Archive; for an archive that still needs millisecond access, choose Glacier Instant Retrieval; to move data by age automatically, use S3 Lifecycle.
Frequently asked questions
How do Glacier Instant Retrieval, Flexible Retrieval and Deep Archive differ?
All three are S3 archive classes: the longer you store, the cheaper it is, but the slower the retrieval. Instant Retrieval returns data in milliseconds, Flexible Retrieval takes minutes to hours depending on the retrieval option, and Deep Archive is the cheapest and takes hours. Choose based on how rarely you need to read the data back.
Are there costs besides storage when using Glacier?
Yes. These classes charge for retrieval and have a minimum storage duration: deleting or moving objects earlier still bills for the full minimum. So they are only cheap when data is truly rarely read; for data accessed more often, S3 Standard-IA or Intelligent-Tiering is usually better.
How do I automatically move old data to Glacier?
Use an S3 Lifecycle rule that transitions objects to a Glacier class after N days and optionally expires them later. Objects stay in the same S3 bucket; for Flexible Retrieval and Deep Archive you must issue a restore request before reading, and the restored copy exists only temporarily for the number of days you choose.
Related services
SAA-C03 study guide
S3 Glacier is covered in these study modules: