Updated 1 Nov 2026 · 7 min read
On this page
- What a storage class is
- The classes side by side
- Frequent access: Standard and Express One Zone
- Unknown access patterns: Intelligent-Tiering
- Infrequent access: Standard-IA and One Zone-IA
- Archive classes and how long retrieval takes
- A worked example: what the minimums cost you
- How this fits your exam
- Frequently asked questions
- Sources
By the end of this page you will be able to pick the right Amazon S3 storage class from a short description of how data is used, and explain the trade-offs behind the choice: how many Availability Zones hold the data, the minimum storage duration and minimum billable object size, and how long it takes to get archived data back. Every storage figure comes from the Amazon S3 user guide as it stood in October 2026. AWS changes these details from time to time, so check the current guide before you rely on a number at work.
What a storage class is
Every object in S3 has a storage class. If you do not specify one when you upload, S3 assigns S3 Standard, the default. You choose a class based on your use case and performance access requirements, and AWS describes the choice as a way to optimise storage costs, performance and availability for your objects. Each class also has an API name, such as STANDARD_IA or DEEP_ARCHIVE.
Durability is not what separates them. The comparison table in the S3 user guide lists 99.999999999% designed durability for every class in it except Reduced Redundancy Storage, a class AWS recommends not using. What changes from class to class is availability, the number of Availability Zones, minimum charges, retrieval fees and how quickly you can read the data.
The classes side by side
| Storage class | Designed for | Availability Zones | Min storage duration | Min billable object size | Designed availability |
|---|---|---|---|---|---|
| S3 Standard | Frequently accessed data, millisecond access | At least 3 | None | None | 99.99% |
| S3 Intelligent-Tiering | Unknown, changing or unpredictable access patterns | At least 3 | None | None | 99.9% |
| S3 Standard-IA | Long-lived, infrequently accessed data, millisecond access | At least 3 | 30 days | 128 KB | 99.9% |
| S3 One Zone-IA | Re-creatable, infrequently accessed data | 1 | 30 days | 128 KB | 99.5% |
| S3 Express One Zone | Single-digit millisecond access for latency-sensitive applications | 1 | None | None | 99.95% |
| S3 Glacier Instant Retrieval | Archive data accessed once a quarter, millisecond access | At least 3 | 90 days | 128 KB | 99.9% |
| S3 Glacier Flexible Retrieval | Archive data accessed once a year, minutes to hours | At least 3 | 90 days | Not applicable (see below) | 99.99% after restore |
| S3 Glacier Deep Archive | Archive data accessed less than once a year, hours | At least 3 | 180 days | Not applicable (see below) | 99.99% after restore |
For Glacier Flexible Retrieval and Glacier Deep Archive, the guide shows the minimum billable size as not applicable and instead notes that each archived object requires 40 KB of additional metadata: 32 KB charged at that class's rate and 8 KB charged at the S3 Standard rate. The "Designed for" wording comes from the comparison table. AWS's separate Glacier page words the recommended access frequency slightly differently: quarterly for Instant Retrieval, semi-annually for Flexible Retrieval and annually for Deep Archive. All classes except One Zone-IA and Express One Zone are designed to be resilient to the physical loss of an Availability Zone.
Frequent access: Standard and Express One Zone
S3 Standard is the default class, for frequently accessed data that needs millisecond access. S3 Express One Zone is a high-performance, single-zone class built for consistent single-digit millisecond access. You choose the one Availability Zone that holds the data, and objects are stored in directory buckets located in Availability Zones. Within that zone the data is stored redundantly on multiple devices. The trade-off is resilience: because it is a single-zone class, it is not designed to survive the physical loss of that zone.
Unknown access patterns: Intelligent-Tiering
S3 Intelligent-Tiering moves objects between access tiers for you based on how they are used. New objects start in the Frequent Access tier. Objects not accessed for 30 consecutive days move to the Infrequent Access tier, and objects not accessed for 90 consecutive days move to the Archive Instant Access tier. You can also activate two optional tiers for data your application can read asynchronously: Archive Access (after at least 90 days without access) and Deep Archive Access (after at least 180 days). Objects in those two tiers must be restored before you can read them.
There are no retrieval fees in Intelligent-Tiering, but there is a monthly monitoring and automation charge per object. Objects smaller than 128 KB are not monitored and always stay in the Frequent Access tier.
Infrequent access: Standard-IA and One Zone-IA
Both classes give millisecond access but charge a retrieval fee, so they suit long-lived data that you read rarely. Both have a 30-day minimum storage duration and a 128 KB minimum billable object size. The difference is where the data lives. Standard-IA stores it across multiple geographically separated Availability Zones, so it is resilient to the loss of a zone. One Zone-IA stores it in only one zone, and AWS says the data is not resilient to the physical loss of that zone from disasters such as earthquakes and floods. One Zone-IA is as durable as Standard-IA but less available and less resilient.
AWS's own recommendation: use Standard-IA for your primary or only copy of data that cannot be re-created, and One Zone-IA for data you can re-create if the zone fails, such as object replicas made with S3 Cross-Region Replication.
Archive classes and how long retrieval takes
AWS's pages label Glacier Instant Retrieval differently. The retrieval options page groups all three Glacier classes as archival storage classes, but the Glacier comparison table marks Instant Retrieval as not archival, and its objects are available immediately with a normal GET, with the same latency and throughput as Standard-IA. Glacier Flexible Retrieval and Glacier Deep Archive are archival: objects are not available for real-time access, and you must first restore them. A restore creates a temporary copy you can read once the request completes. You choose a retrieval tier when you restore.
| Retrieval tier | Glacier Flexible Retrieval | Glacier Deep Archive |
|---|---|---|
| Expedited | 1 to 5 minutes (objects under 250 MB) | Not available |
| Standard | 3 to 5 hours | Within 12 hours |
| Bulk | 5 to 12 hours | Within 48 hours |
Standard is the default when a restore request does not name a tier. The Standard times shown are for restores made without S3 Batch Operations. Expedited retrievals are subject to demand: without purchased provisioned capacity, they might not be accepted during periods of high demand. The same times apply to the Intelligent-Tiering Archive Access and Deep Archive Access tiers respectively. All of these are "typically" figures in AWS's wording, not guarantees, and restores of very large datasets can take longer.
A worked example: what the minimums cost you
Minimums matter more than they look. Suppose you upload a 50 KB object to Standard-IA and delete it after 10 days.
- Size: the object is under 128 KB, so S3 charges you as if it were 128 KB.
- Duration: you are charged the normal storage charge for the 10 days it existed, plus a pro-rated charge for the remaining 20 days of the 30-day minimum. You pay for 30 days in total.
S3 Standard has no minimum storage duration and no minimum billable size, so the same object there is billed as 50 KB for the 10 days it existed. Now apply that to log files under 128 KB that are written once, read heavily for a week and deleted after two weeks. Standard-IA and One Zone-IA are the wrong choice on both counts, and the heavy reads would add per-GB retrieval fees on top. A backup that must be kept for years, is read less than once a year and can wait hours for a restore points to Glacier Deep Archive. The same backup with a requirement for millisecond reads about once a quarter points to Glacier Instant Retrieval instead.
How this fits your exam
AWS SAA-C03. Content Domain 4, Design Cost-Optimized Architectures, is 20% of the scored content. Task 4.1, "Design cost-optimized storage solutions", lists knowledge of data lifecycles and "Storage tiering (for example, cold tiering for object storage)", and skills including "Managing S3 object lifecycles" and "Selecting the appropriate storage tier". Task 2.2, "Design highly available and/or fault-tolerant architectures", also lists "Storage options and characteristics (for example, durability, replication)". Our SAA-C03 study guide covers the rest of the exam.
AWS CLF-C02. Content Domain 3, Cloud Technology and Services, is 34% of the scored content. Task Statement 3.6, "Identify AWS storage services", includes the skills "Recognizing the differences in Amazon S3 storage classes" and "Understanding use cases for lifecycle policies". See the CLF-C02 study guide for the other domains.
Frequently asked questions
Is One Zone-IA less durable than Standard-IA?
No. AWS lists the same designed durability for both. One Zone-IA is less available (99.5% against 99.9%) and less resilient, because it keeps data in a single Availability Zone and is not designed to survive the physical loss of that zone.
Can I read an object in Glacier Flexible Retrieval straight away?
No. Objects in Glacier Flexible Retrieval and Glacier Deep Archive are archived, so you first restore them, which creates a temporary copy you can read. Of the three Glacier classes, only Glacier Instant Retrieval gives immediate access with a normal GET.
Which class should I use if I do not know how often data will be read?
AWS describes S3 Intelligent-Tiering as the ideal storage class for data with unknown or changing access patterns. It moves objects between tiers automatically and has no retrieval fees, though objects under 128 KB are not monitored and stay in the Frequent Access tier.
Sources
- Understanding and managing Amazon S3 storage classes
- Understanding S3 Glacier storage classes for long-term data storage
- Understanding archive retrieval options
- High performance workloads
- Content Domain 4: Design Cost-Optimized Architectures
- Content Domain 3: Cloud Technology and Services
Check your storage class choices with the free SAA-C03 practice questions, then read the full SAA-C03 study guide.