Amazon S3 Glacier
Amazon S3 Glacier is the family of Amazon S3 storage classes for low-cost archival data, ranging from millisecond access to deep-archive retrieval measured in hours. The older Amazon Glacier vault/archive service is a separate legacy API model.
AI Practitioner focus
- Glacier storage classes preserve infrequently accessed datasets, model artifacts, audit evidence, and logs for long-term retention at low storage cost with retrieval delay/fees.
- Do not use archival classes for active training or low-latency RAG. Match retention and deletion to legal, governance, reproducibility, and data-minimization requirements.
Key points
- Glacier storage classes inherit Amazon S3’s 99.999999999% (11 nines) durability design and store data redundantly across multiple Availability Zones.
- S3 Glacier Instant Retrieval is for rarely accessed archives that still require millisecond retrieval; it has a 90-day minimum storage duration.
- S3 Glacier Flexible Retrieval is for archives accessed roughly once or twice per year, with three retrieval tiers: Expedited (typically 1-5 minutes), Standard (typically 3-5 hours; the default tier if none is specified), and Bulk (typically 5-12 hours, the lowest-cost option); it has a 90-day minimum storage duration.
- S3 Glacier Deep Archive is the lowest-cost S3 storage class, designed for data accessed less than annually; it offers only Standard (typically within 12 hours) and Bulk (typically within 48 hours) retrieval — Expedited retrieval is not available for Deep Archive — and the minimum storage duration is 180 days.
- Flexible Retrieval and Deep Archive objects must be restored to a temporary accessible copy before normal reads; Instant Retrieval does not require restoration.
- Use lifecycle rules to transition objects into an archive class and expire them according to retention and deletion policy.
- S3 server-side encryption options include SSE-S3 and SSE-KMS; access remains controlled through IAM and bucket/resource policies.
Pricing
- Cheapest S3 storage class.