CloudCostCompareObject storage classes and lifecycle rules, explained
Object storage holds files cheaply and durably. Every cloud offers it in several classes, with different prices for keeping data and for reading it. Choosing the right class is the main lever on a large storage bill.
The pattern
| Class idea | Price to keep | Price to read | Use for |
|---|---|---|---|
| Hot or standard | Highest | None or lowest | Data you read often |
| Infrequent access | Lower | A per-gigabyte retrieval fee | Data read a few times a year |
| Archive | Lowest | Higher retrieval fee, may take minutes to hours | Compliance copies, old raw data |
The product names are Amazon S3, Azure Blob Storage, Google Cloud Storage and Oracle Object Storage, and each names its classes differently. Our storage catalog lists the per-gigabyte prices for the S3 classes we have read, including the volume tiers within a class.
Charges beyond the monthly price
The monthly price per gigabyte is only part of the picture.
- Retrieval fees. Colder classes charge per gigabyte read. A cheap archive class becomes expensive if you read it often.
- Minimum storage duration. Many colder classes bill a minimum number of days, for example thirty or ninety, even if you delete earlier.
- Minimum object size. Some classes bill a minimum size per object, which makes millions of tiny files costly.
- Requests. Reads and writes are billed per thousand or per ten thousand, and the rates differ by class.
- Transitions. Moving objects between classes can carry a per-object charge.
- Data transfer. Reading data out to the internet or another region is egress.
Our catalog covers the per-gigabyte storage prices only, and says so on the page. Requests, retrieval and transfer need the vendor’s own page.
Lifecycle rules
A lifecycle rule moves or deletes objects automatically by age. A typical policy:
- Keep new data in the hot class for thirty days.
- Move it to infrequent access after that.
- Move it to archive after a year.
- Delete it after the retention period your rules require.
Rules can also clean up incomplete uploads and old object versions, two common sources of invisible storage.
Intelligent tiering
Some classes move objects between tiers for you, based on access, for a small monitoring charge per object. They suit data with unpredictable access and large objects. They suit tiny objects less, because of the per-object fee.
Setting rules safely
- Check how your data is actually read before choosing. A bucket you scan every day does not belong in an archive class.
- Beware of analytics on cold data. A query engine reading archived files can trigger large retrieval charges. See data lake cost basics.
- Test a rule on a small prefix first.
- Turn on versioning only with a rule to expire old versions.
- Remember legal holds and compliance before any deletion rule.
Related reading
How long to keep sensor data and where is covered in object storage for sensor archives. Backups are covered in backup and retention costs, and the other kind of storage, attached to servers, is in block storage volume types.