Storage fundamentals
Storage class
A storage class is a named level of service that an object storage system applies to an object, defining how the object is protected, how quickly it can be read and what it costs to keep. The object's name and the API used to reach it stay the same across classes. What changes is the media, redundancy and retrieval behavior underneath.
What distinguishes one storage class from another?
Classes trade access speed against cost. Public cloud services publish a menu, and on-premises platforms define their own.
| Class type | Access time | Typical use | Trade-off |
|---|---|---|---|
| Standard or hot | Immediate | Active data, frequent reads | Highest price per terabyte |
| Infrequent access | Immediate | Backups, older data read occasionally | Lower storage price, retrieval fee, minimum duration |
| Archive | Minutes to hours | Compliance copies, long-term retention | Lowest price, restore step before reading |
| Reduced redundancy or single-site | Immediate | Reproducible data | Lower durability or availability |
How is a storage class assigned?
A class is set per object when it is written, through a header on the request, or inherited from a default. It can be changed later by copying the object onto itself with a new class, or automatically by lifecycle rules that transition objects after a number of days. The mechanism is described under storage lifecycle management, and the general idea of matching data to media under storage tiering.
What should be checked before using a colder class?
Three conditions catch people out. Minimum storage durations mean an object deleted or changed early is billed as if it had stayed. Retrieval fees are charged per gigabyte read, so a large restore from a cold class can cost more than the storage saved. And archive classes need a restore request before data is readable, which adds hours to recovery. Small objects can also be charged at a minimum billable size. These conditions suit data that is written once and rarely touched, the pattern described under cold storage.
What do storage classes mean on-premises?
On a platform an organization operates, a storage class maps to choices it controls: the erasure coding or replication policy, the number of sites, and the type of drives. There are no retrieval fees, and the difference between classes is the capacity overhead and hardware behind each. Offering a small set of classes with clear names lets application teams choose a level without needing to know the mechanics. Storage service levels covers how to define them.
Frequently asked questions
Does changing storage class change the object's address?
No. The bucket and key stay the same. Only the service level behind them changes.
Can one bucket hold objects in different classes?
Yes. Storage class is a property of each object, not of the bucket.
Is the cheapest class always the most economical?
Not if the data is read or deleted early. Retrieval fees and minimum durations can outweigh the lower storage price.
Related terms
- Storage tiering: matching data to media by access pattern.
- Hot storage: the class for active data.
- Cold storage: the class for rarely read data.
- Storage lifecycle management: rules that move objects between classes.
- Cloud archive storage: the archive end of the range.