Durability, quotas and billing

Durability

Ceph spreads your data across many disks and several hosts, and protects it with one of two redundancy schemes depending on the storage type and pool:

• 3-way replication keeps three full copies of every object on different hosts. It is fast and rebuilds quickly, at the cost of using three times the raw capacity.
• Erasure coding splits data into shards plus parity, so it survives failures with much less storage overhead than three full copies. It is used where capacity efficiency matters more than the lowest possible latency.

Storage	Protection
Block storage, type fast	3-way replication
Block storage, types balanced and bulk	Erasure coding, 6+2
Object storage (S3)	Erasure coding

With erasure coding 6+2 the data is split into 6 data shards plus 2 parity shards, stored on 8 different hosts. Any two of those hosts can fail at the same time without data loss, and the raw capacity used is about 1.3 times the data, compared to 3 times for replication. Small random writes are slightly slower than on replicated storage, so choose fast for latency-sensitive workloads such as databases. See Block storage for the volume types.

Either way, if a disk or a whole node fails, the data is still served from the remaining copies or shards and Ceph rebuilds the missing redundancy automatically. You do not need to configure RAID or replication yourself.

This protects against hardware failure. It does not protect against:

• Deleting data yourself (the delete is replicated immediately).
• An application or ransomware overwriting or encrypting your files.
• Deleting a volume together with all of its snapshots.

For those cases you need a backup that lives outside the cluster, or at least object versioning with a lifecycle policy you control.

What is not backed up by default

Object (S3), block, and filesystem storage are not backed up by Sunet unless you have specifically ordered a backup service. Plan your own strategy:

• For S3, enable versioning and lifecycle rules, or copy critical buckets to another location with rclone or aws s3 sync.
• For OpenStack volumes, take and export snapshots, or back up at the application level (for example database dumps to S3).
• For Kubernetes, back up persistent volumes and cluster state, for example with Velero.

Capacity limits

The amount of object storage (S3) in your order is a soft commercial limit, not a technical quota. Capacity is consumed dynamically as you upload objects, and exceeding the ordered amount does not automatically stop writes. Usage is metered, and Sunet will contact you about increasing your order if you use more storage than it includes.

Other storage services, such as block storage, may enforce technical project quotas. Requests to create or extend a resource fail when such a quota is reached. Customer administrators manage project quotas from the project's Edit page in the Customer Portal.

Billing

Storage is billed on usage per month. Object storage (S3) is billed per gigabyte stored, with lower rates for longer commitment terms; block storage is billed per gigabyte by tier. Usage is metered and aggregated across all of an organisation's projects.

For current, authoritative pricing, see the official Sunet VDC price list.