S3 Storage Classes

• Replication and resiliency (multi-AZ vs single-AZ)
• Retrieval latency and retrieval fees
• Minimum storage duration and minimum billable object size
• Typical use cases and cost trade-offs

​

S3 Standard (default)

• Default storage class when none is specified during upload.
• Redundantly stores objects across multiple Availability Zones (designed to sustain two simultaneous AZ failures).
• 11 nines of durability (99.999999999%) and low (millisecond) GET latency.
• Charged per gigabyte per month for storage, plus egress charges on retrieval.
• No retrieval fee, no minimum storage duration, and no minimum object size billing.
• Objects can be served publicly via bucket policy/ACL or presigned URLs.

​

S3 Standard‑Infrequent Access (Standard‑IA)

• For data accessed less frequently but requiring rapid access when requested.
• Multi-AZ redundancy and similar durability to Standard.
• Lower storage cost per GB than S3 Standard, but retrievals incur an additional retrieval fee (plus egress).
• Minimum storage duration: 30 days. Minimum billable object size: 128 KB (objects smaller than 128 KB are billed as 128 KB).
• Low-latency (millisecond) access; can be made public if required.

​

S3 One Zone‑Infrequent Access (One Zone‑IA)

• Same access and retrieval characteristics as Standard‑IA but stores objects in a single Availability Zone (no multi‑AZ redundancy).
• Lower storage cost than Standard‑IA because it does not provide AZ-level redundancy.
• Retrieval fee applies; minimum storage duration: 30 days; minimum billable object size: 128 KB.
• Low-latency access while the AZ is available; however, the single-AZ design increases risk if that AZ becomes unavailable.
• Objects can be made public when accessible.

​

The S3 Glacier family (archival storage)

​

S3 Glacier Instant Retrieval

• For long-lived archive data that is rarely accessed but requires immediate (millisecond) access when requested.
• Multi-AZ durability (11 nines).
• Very low storage cost per GB compared with Standard and Standard‑IA.
• Retrieval fees apply. Minimum storage duration: 90 days. Minimum billable object size: 128 KB.
• No restore step required — objects are available for GET immediately.

​

S3 Glacier Flexible Retrieval

• Archival class for rarely accessed data where immediate access is not required.
• Objects must be restored (a retrieval operation) before they are available for GET.
• Multi-AZ durability like other multi-AZ tiers.
• Lower storage cost per GB than non-archival tiers.
• Minimum storage duration: 90 days. Minimum billable object size: 40 KB.
• Retrieval options (cost and time trade-offs):
  • Expedited: ~1–5 minutes
  • Standard: ~3–5 hours
  • Bulk: ~5–12 hours
• Restored data is temporarily available (often staged in a different access tier such as Standard‑IA) for the chosen restoration period.

​

S3 Glacier Deep Archive

• Lowest-cost S3 storage class for long-term retention of rarely accessed data.
• Objects must be restored before access; retrievals are the slowest among Glacier classes.
• Multi-AZ durability.
• Minimum storage duration: 180 days. Minimum billable object size: 40 KB.
• Retrieval options:
  • Standard: typically accessible within ~12 hours
  • Bulk: typically accessible within ~48 hours
• Retrieval fees and egress charges apply. Use slower, cheaper retrieval options (Bulk) when possible.
• Restored data may be staged temporarily in a faster access tier for the restoration window.

​

S3 Intelligent‑Tiering

• If you can’t predict the optimal storage class per object, Intelligent‑Tiering automatically moves objects between access tiers to reduce costs.
• Monitors access patterns and transitions objects to the most cost‑effective tier.
• Small monitoring/automation fee in addition to the storage fee for the tier the object currently occupies (charged per 1,000 objects).
• Supports multiple access tiers (frequent, infrequent) and optional Archive/Deep Archive tiers (these introduce archival retrieval characteristics).

​

Quick decision flow

• Need immediate access?
  • Frequently accessed → S3 Standard
  • Infrequently accessed but immediate when needed → Standard‑IA or One Zone‑IA (choose One Zone‑IA only if you can tolerate single‑AZ risk)
  • Almost never but must be immediately available → Glacier Instant Retrieval
• Do not need immediate access:
  • Rarely accessed, can wait hours → Glacier Flexible Retrieval
  • Very rarely accessed, long‑term retention with minimal cost → Glacier Deep Archive
• Unsure / unpredictable access patterns → S3 Intelligent‑Tiering

​

Storage classes at-a-glance

​

Summary

• Storage classes let you balance access latency, resiliency, and cost to match workload requirements.
• Set the storage class when uploading an object, or transition objects later with lifecycle rules or by copying the object to the same key with a different storage class.

​

Examples: specifying a storage class on upload

• S3 REST PUT with header (example using curl — header shown as backticks per MDX rules):

curl -X PUT "https://my-bucket.s3.amazonaws.com/my-object" \
  -H "x-amz-storage-class: STANDARD_IA" \
  -T "./my-object"

• AWS CLI example:

aws s3 cp ./my-object s3://my-bucket/my-object --storage-class STANDARD_IA

• To change the storage class of an existing object, copy it to the same key with the desired storage class (CLI):

aws s3 cp s3://my-bucket/my-object s3://my-bucket/my-object --storage-class GLACIER_IR