Questions? Feedback? powered by Olark live chat software

Running on a software stack that has at its base a custom-build of the FreeBSD operating system, HybridCluster achieves its Rapid-restore Data Vault by taking advantage of ZFS’ snapshot capabilities. This allows for the creation of read-only copies of file systems and the subsequent ability to revert a live, writable file system back to a previous version.

Snapshots are different to conventional backups in that they are only a mapping of meta-information about a file system – that is, a snapshot initially consumes no disk space in its own right, nor makes actual copies of any files contained within a volume (prior to the point of replication). Disk space resources are only consumed by a snapshot in response to a change of data, protecting the blocks that were used for an altered file and preventing those blocks being overwritten by new or modified data.

While this provides excellent protection for the integrity of data within a file system in respect of accidental loss or change on the part of a user, there is of course a degree of balance to be struck between the need to keep available previous versions of data and the amount of space consumed on disk; frequent changes to files can cause a rapid accumulation of disk usage, even up to an order of magnitude many times greater than the file system that’s live.

To this end, HybridCluster provides two solutions straight out of the box – the first being that snapshots are included in the quota of storage you allocate to your customers, meaning they are included in the measurement of disk space consumed by a user account. The second is that our platform will automatically prune snapshots, freeing up blocks for reuse when a snapshot is no longer needed. A balance is therefore achieved by ensuring that your customers will always be able to rollback any changes that have taken place to their data within the last hour (regardless of frequency), in addition to a sliding scale providing hourly snapshots for the last 24 hours, daily for the last month and weekly for a file system’s entire history.

— Back to top —