A practical storage-class mental model
S3 Standard
Default choice for frequently accessed data requiring millisecond access and multi-AZ resilience.
S3 Intelligent-Tiering
For unknown or changing access patterns. Eligible objects move automatically among access tiers; monitoring and automation fees apply.
Standard-IA and One Zone-IA
For long-lived, infrequently accessed objects needing millisecond access. Standard-IA is multi-AZ; One Zone-IA is for reproducible or secondary data.
S3 Glacier classes
Instant Retrieval provides millisecond access. Flexible Retrieval and Deep Archive require a restore and suit progressively rarer access.
Lifecycle rules automate known data aging
Lifecycle transition actions move eligible objects to another storage class. Expiration actions delete eligible objects. Rules can filter by prefix, tags, object size or combinations, and they apply to existing as well as new objects.
Current small-object behavior: lifecycle configurations created or modified under the current default do not transition objects smaller than 128 KB unless an object-size filter explicitly permits it. For many tiny objects, transition-request and minimum-size costs can outweigh storage savings.
Also account for minimum storage-duration charges. Deleting or transitioning an IA or archive object too early can still incur charges for the remaining minimum duration.
Scenario-based architecture decisions
Application logs: hot for 30 days, compliance archive for seven years
Keep new logs in S3 Standard, transition older objects to an appropriate archive class after the active investigation period, and expire them only after the required retention period. Confirm whether auditors require immediate access; that determines Instant Retrieval versus a restore-based class.
User media with unpredictable popularity
Use Intelligent-Tiering when objects may become hot again and the access pattern cannot be predicted confidently. This avoids maintaining fixed age-based transitions that do not reflect actual reads.
Re-creatable thumbnails accessed infrequently
One Zone-IA can be considered because the thumbnails can be regenerated. Do not select it for the only copy of irreplaceable source media that must survive an Availability Zone loss.
Common SAA-C03 mistakes
- Choosing the lowest storage price without considering retrieval charges, transition requests or restore time.
- Using One Zone-IA for the only copy of data that cannot be recreated.
- Assuming every Glacier class requires the same retrieval workflow.
- Using a fixed lifecycle schedule when the access pattern is unknown and Intelligent-Tiering fits better.
- Ignoring minimum storage duration and small-object considerations.
- Confusing lifecycle expiration with an IAM-authorized user deletion; lifecycle rules continue independently of bucket-policy denies.
Fast decision rule
Known aging pattern: lifecycle rules. Unknown or changing pattern: Intelligent-Tiering. Irreplaceable infrequent data: prefer multi-AZ Standard-IA over One Zone-IA. Archive: choose by required retrieval time.
Official AWS references
Frequently asked questions
When should I choose S3 Intelligent-Tiering?
Choose it when access patterns are unknown, changing or difficult to predict. It automatically moves eligible objects among access tiers and charges a small monitoring and automation fee per object.
What is the key difference between S3 Standard-IA and One Zone-IA?
Standard-IA stores data across multiple Availability Zones. One Zone-IA stores data in one Availability Zone, so it is best for reproducible data or secondary copies that can tolerate the loss of that zone.
Do S3 Glacier objects require a restore?
S3 Glacier Flexible Retrieval and Deep Archive objects must be restored before access. S3 Glacier Instant Retrieval supports millisecond access without that restore workflow.
Do lifecycle rules apply to existing objects?
Yes. When a lifecycle configuration is added, eligible rules apply to existing objects as well as new objects, based on object age and the rule filters.