Vault 2015

Backup-Restore feels like the unloved step-child of IT: neither cool nor wanted, everybody needs and suffers from implementing it.

For a long time it was sufficient to implement three approaches: reading the filesystem, working with LVs or approaching the block layer.

Today, xfs provides freezing and exporting, btrfs and LVM do CoW and snapshots, btrfs implements send-receive (a potential hook for backup and restore?). Different for scale out filesystems (such as Ceph): are they only backend for backup or do they need to be backed-up as well? How? There is still no generic approach to synchronize active "freezing" of an application with checkpoint-restart on the storage level.

We will discuss some of the existing backup/restore technologies, investigate the differences between the local and distributed cases and explore possibilties of introducing a common API.