Table of contents
Backup storage tiers
Backup storage tiers are classes of storage that hold backup copies at different balances of cost, access latency and retrieval cost. Recent restore points usually sit on flash or fast disk, older ones on high-capacity disk or object storage, and long-term copies on tape or a cloud archive class.
How do the tiers compare?
Three tiers cover most backup designs. They differ on what they cost to keep, how soon the first byte can be read, and what it costs to read data back.
| Tier | Typical media | Access latency | Cost to keep | Cost to read back |
|---|---|---|---|---|
| Performance | Flash or fast disk | Immediate | Highest per terabyte | No per-read charge |
| Capacity | High-density disk or object storage | Seconds | Lower per terabyte | None on premises; request and egress fees in public cloud |
| Archive | Tape or a cloud archive class | Minutes to hours | Lowest per terabyte | Retrieval fees and minimum storage periods in public cloud; drive and handling time for tape |
The backup software applies the rules that move restore points from one tier to the next.
How do on-premises and public cloud tiers differ?
On premises, a tier costs hardware, power and space, and reads carry no per-request charge. An S3-compatible object store in the data center can act as the capacity tier and returns every read at the same speed and cost. Public cloud storage classes price access separately. Archive classes add a retrieval fee per gigabyte or per request, a delay before the first byte is readable, and a minimum storage period with charges for early deletion. A large restore from any cloud class also pays egress fees on every byte read out, so a cloud tier that is cheap to keep can be slow and expensive to restore from.
What does tiering by age get wrong?
Calendar age is the usual trigger: anything older than 14 days moves to the capacity tier, anything older than 90 days moves to archive. The trigger is easy to configure and only loosely related to what people actually restore. If restores routinely reach back three weeks, for example because corruption is noticed late, data that left fast storage at day 14 turns each of those restores into a slow, sometimes billed, retrieval. Boundaries set from the real restore history rarely match round numbers.
Do retention locks follow data between tiers?
A lock is a property of the stored object. When a restore point moves tiers, the move is a copy followed by a delete, and the immutable guarantee holds only if the lock date is carried to the new location and the source copy is not removed through a path the lock does not cover. A policy that drops the lock at the boundary leaves the oldest backups the easiest to delete.
Frequently asked questions
Is the archive tier the same as cold storage?
Archive is a backup term for the lowest-cost tier of a retention chain. Cold storage describes any data that is rarely accessed. A backup archive tier is one place cold data lives.
How many backup tiers does a design need?
Two is common: fast storage for recent restore points and a lower-cost tier for the rest. A third archive tier appears when long-term retention requirements or very large histories make the capacity tier expensive.
Can backups be restored directly from the capacity tier?
Usually yes, at a lower speed than the performance tier. Some backup software also needs to read the restore point back into the performance tier first, which adds copy time before the restore starts.
Related terms
- Backup target: the system a tier is built from.
- Retention policy design: how long each restore point lives and where.
- Restore throughput: how fast each tier hands data back.
- Storage tiering: the general practice of placing data by access pattern.
- Cloud archive storage: the slowest, cheapest end of the scale.
- Egress fees: the charge for reading data back out of a cloud.