Weekly Active Fulls to improve restore speeds: good practice or unnecessary?

[StoopidMonkey]

On the few occasions we've had to restore full backups the speeds have not been great, usually averaging around 20-25MB/s for most of the job. I've read that a possible cause is the amount of Random I/O involved in restoring incremental jobs and/or those using synthetic fulls (repository is a RAID10 of 7.2K disks). Restores from Active Full backups seem to be a bit faster and I'm guessing that's because the most of the data is sequential vs random. I recently changed my backups to perform one Active Full per week but I wanted to check in to see if that's really a good idea, or if my understanding of what Synthetic Fulls are is correct. When using Fast Cloning and ReFS, aren't they really a collection of pointers that still result in lots of Random I/O to restore from, just maybe with the benefit of not having to process dozens of incremental backups to achieve the final result?

[HannesK]

Hello,
correct, with REFS / XFS synthetic fulls, it's still random IO.

So active full might improve speed. How much, is hard to say, and whether it's worth the additional load and disk space. The easiest way to find out for your environment is to change the settings. 20-25MByte/s sounds very slow in general, but I don't know your setup / environment.

Best regards,
Hannes