Amazon S3 Storage Classes: How to Choose the Right Tier

Choosing an Amazon S3 storage class is a cost and access decision. The right question is not simply “which class is cheapest?” It is: how often will the object be read, how quickly must it be available, and can the data be recreated if an Availability Zone is lost?

This is Part 13 of the AWS Cloud Practitioner Study Notes series. Part 11 introduced S3 buckets, objects, and core features, while Part 12 compared object storage with block and file storage. This note focuses on selecting an S3 class for a workload.

Choose by access pattern

The storage classes form a useful decision spectrum. As data becomes less frequently accessed, AWS offers lower storage rates, but retrieval fees, minimum storage durations, or restore delays may become more important.

Storage class	Best access pattern	Retrieval experience	Main decision cue
S3 Standard	Frequent access	Milliseconds	Active application data
S3 Intelligent-Tiering	Unknown or changing access	Milliseconds in active tiers	Let S3 optimise automatically
S3 Standard-IA	Infrequent access	Milliseconds	Important data that still needs immediate access
S3 One Zone-IA	Infrequent, recreatable data	Milliseconds	A single Availability Zone is acceptable
S3 Glacier Instant Retrieval	Rare archive access	Milliseconds	Archive data that must be immediately available
S3 Glacier Flexible Retrieval	Rare archive access	Minutes to hours	Retrieval can be asynchronous
S3 Glacier Deep Archive	Very rare, long-term retention	Usually hours	Lowest-cost long-term archive

The table is a study shortcut, not a complete pricing model. Actual cost also depends on object size, request volume, data retrieval, transition requests, and how long each object remains in a class.

S3 Standard: frequent access

S3 Standard is the general-purpose class and the sensible starting point for objects that an application reads regularly. It provides millisecond access and is designed for data stored redundantly across at least three Availability Zones.

Typical examples include:

• Website images and other active static assets
• User uploads that are frequently viewed or downloaded
• Mobile application content
• Working documents and active datasets

For an exam question that emphasises frequent access and low-latency retrieval, S3 Standard is usually the intended answer. It has a higher storage rate than colder classes, but avoiding retrieval fees and archive restores can make it the better total-cost choice for active data.

S3 Intelligent-Tiering: unknown or changing access

S3 Intelligent-Tiering is designed for objects whose access pattern is unknown, unpredictable, or likely to change. S3 monitors access and moves objects between access tiers automatically while the object remains in the Intelligent-Tiering storage class.

The automatic tiers include:

• Frequent Access: the starting tier for new objects and objects that continue to be used.
• Infrequent Access: objects that have not been accessed for 30 consecutive days.
• Archive Instant Access: objects that have not been accessed for 90 consecutive days, while retaining millisecond access.

Optional Archive Access and Deep Archive Access tiers can be enabled when asynchronous retrieval is acceptable. Those tiers reduce storage cost further, but archived objects must be restored before use and retrieval can take minutes to hours.

Intelligent-Tiering has a small per-object monitoring and automation charge. It is a strong fit when manually predicting and maintaining lifecycle transitions would be difficult. It is less compelling when access behaviour is already well understood and a fixed class is clearly cheaper.

S3 Standard-IA: infrequent but immediate access

S3 Standard-IA, where IA means infrequent access, is for long-lived data that is read less often but must still be available in milliseconds when requested. It is stored redundantly across multiple Availability Zones and is suitable for a primary or only copy that cannot easily be recreated.

Common examples include:

• Monthly backups that may need a fast restore
• Disaster-recovery data
• Older invoices and business records
• Data that is retained for a long time but rarely opened

Standard-IA has retrieval charges and a 30-day minimum storage duration. AWS documentation also notes a 128 KB minimum billable object size for this class. Therefore, moving many small or short-lived objects into Standard-IA may cost more than expected.

S3 One Zone-IA: infrequent and recreatable data

S3 One Zone-IA provides millisecond access for infrequently accessed data, but stores it redundantly within one Availability Zone. This makes it less expensive than Standard-IA, while also making it less available and less resilient to the physical loss of that Availability Zone.

Use it when the data can be recreated or another copy exists, for example:

• Temporary backups
• Secondary copies created by a replication process
• Derived thumbnails or transformed datasets
• Data that can be regenerated from a source system

Do not use One Zone-IA as the only copy of irreplaceable data merely because its storage price is lower. It also has retrieval charges and a 30-day minimum storage duration.

Glacier classes: archive with different retrieval expectations

The S3 Glacier classes are still S3 storage classes. Objects remain in Amazon S3 and are managed through S3 APIs, lifecycle rules, and restore operations where applicable.

S3 Glacier Instant Retrieval

Glacier Instant Retrieval is for long-lived archive data that is rarely accessed but must be available in milliseconds when needed. It is a good fit for archives such as medical images, legal records, or media assets that are retained for a long time but occasionally viewed immediately.

It has retrieval charges and a 90-day minimum storage duration. The phrase “archive” does not automatically mean “hours to retrieve”; when the question says archive data needs instant access, Glacier Instant Retrieval is the key clue.

S3 Glacier Flexible Retrieval

Glacier Flexible Retrieval is for archives where retrieval can be asynchronous. Depending on the retrieval option, restoring an object can take minutes to hours. Archived objects are not available in real time; an application must restore a temporary copy before reading them.

Typical use cases include:

• Compliance archives
• Historical financial records
• Older audit logs
• Backups that are rarely restored

This class has a 90-day minimum storage duration and retrieval charges. It is appropriate when lower storage cost matters more than immediate availability.

S3 Glacier Deep Archive

Glacier Deep Archive is designed for data that is almost never accessed and must be retained for years. Standard retrieval is measured in hours, and the class has a 180-day minimum storage duration.

Examples include:

• Seven-year regulatory retention
• Government or historical records
• Long-term tax documents
• Compliance data that is rarely retrieved

For an exam question containing “lowest-cost long-term archive” or “retention for several years,” Glacier Deep Archive is usually the intended answer. Confirm that the business can wait for restoration and can absorb the minimum-duration and retrieval constraints.

Lifecycle rules automate the movement

S3 Lifecycle lets you define transition and expiration actions for groups of objects. A common policy might look like this:

1. Upload new application logs to S3 Standard.
2. Transition older logs to Standard-IA after their active analysis period.
3. Transition them to Glacier Flexible Retrieval when access becomes rare.
4. Expire them after the retention period ends.

The exact timing depends on the workload. For example, a seven-year insurance-record policy could keep new claims in Standard, move them to Standard-IA after 90 days, archive them after one year, and expire them after seven years.

Before creating a rule, check:

• Whether the object can be recreated or needs multi-AZ resilience
• Whether users need milliseconds, minutes, or hours for retrieval
• Minimum storage durations and minimum billable object sizes
• Transition request costs and the number of objects affected
• Versioning, because previous object versions may need their own lifecycle actions

Lifecycle transitions are not automatically free just because the destination storage rate is lower. A transition that happens too soon, or affects many small objects, can undermine the expected saving.

Common CLF-C02 exam cues

Use these phrases as a quick matching exercise:

• Frequently accessed: S3 Standard
• Unknown or changing access pattern: S3 Intelligent-Tiering
• Infrequent access with immediate retrieval and multi-AZ resilience: S3 Standard-IA
• Infrequent, recreatable data in one Availability Zone: S3 One Zone-IA
• Archive with millisecond retrieval: S3 Glacier Instant Retrieval
• Archive with minutes-to-hours retrieval: S3 Glacier Flexible Retrieval
• Lowest-cost, years-long retention: S3 Glacier Deep Archive
• Automatically transition or delete objects by age: S3 Lifecycle

The exam may describe the same requirement using different words. Focus on the combination of access frequency, retrieval time, resilience, and retention period rather than memorising only a product name.

Conclusion

S3 Standard fits active data. Intelligent-Tiering fits uncertain access. Standard-IA and One Zone-IA fit different kinds of infrequent access, with One Zone-IA trading resilience for lower storage cost. The Glacier classes are for archives, and their key difference is how quickly data must be restored.

The practical choice is usually a policy rather than a single class: upload data to a class that matches its current access pattern, then use Lifecycle rules to transition or expire it as its value and usage change.

Sources

• Understanding and managing Amazon S3 storage classes
• How S3 Intelligent-Tiering works
• Transitioning objects using Amazon S3 Lifecycle
• Data protection in Amazon S3