Questions about ReFS & general backup to tape setup

[MaxDes]

Good day folks!

So a quick rundown:
Running one backup server - Server 2016 64GB ram Xeon E5-2623 v3, array connected via SAS perc840 (8, 8TB drives) plus another 40TB array running off a perc740 (I think).

We have a mix of physical and virtual (vmware) and Oracle data.

Ive read some things about making sure RAID is Veeam optimized (256k stripe/block vs 64k) Which I plan to do, but I have some other questions in regard to ReFS, my volume and repository setup, and backing up to tape.

We, currently, do backups of the weeks full/incrementals from disk to tape over the weekend starting Friday evening. We have off site retention policies, which require a full backup done weekly and moved off site.

How does this work with ReFS? It seems you lose all of the space saving when you dump that to tape right? So would deduplication be a better choice?
Also when it comes to ReFS, does it look at just the backup data of the server being backed up to create the pointers and save space? Or is it across the board like.... this server also has all of these identical files, so ReFS just creates a pointer vs making a copy of that data?

Is it best to have one large volume on your raid array and then just create separate repositories in order to organize and separate your backup job types (oracle vs VM vs physical machine vs workstation?) Or create a volume for each type of backup? Does that matter at all?

I'm sure more will come up, but maybe someone will want to start a dialogue about this stuff with me?

Thank you

[Dima P.]

Hello and welcome to the community Marc!

How does this work with ReFS? It seems you lose all of the space saving when you dump that to tape right? So would deduplication be a better choice?

Data gets 're-hydrated' in during tape outs, so yes you don't get any benefits from disk storage savings while writing data to tape.

Also when it comes to ReFS, does it look at just the backup data of the server being backed up to create the pointers and save space? Or is it across the board like.... this server also has all of these identical files, so ReFS just creates a pointer vs making a copy of that data?

Tape job will request all the needed blocks to make backup files self-sufficient (otherwise you would need your original repository to restore anything from tape )

Is it best to have one large volume on your raid array and then just create separate repositories in order to organize and separate your backup job types (oracle vs VM vs physical machine vs workstation?) Or create a volume for each type of backup? Does that matter at all?

Since tape jobs are reading data from the repository they do rely on storage I/O. RAID will speed up the data read for sure, however, in terms of organization (volumes/folders) it's fully up to you.

[MaxDes]

Thanks for the welcome!!!

Does rehydration slow down the backup to tape job?

So the rehydration happens, that makes sense, is it possible to apply deduplication to the backup to tape job coming from an ReFS volume? (for instance, if the backup job is backing up an entire repository as opposed to one server, because obviously doesn't seem like there would be anything to deduplicate if its just the one server.

Also due to it's nature, I am guessing Refs doesn't much help with oracle data either? Especially encrypted data? (Kinda talking myself through also as I read and write, so thanks for bearing with me. )

If I have a job and it does a weekly full and then incrementals the other days is that two separate jobs?

How would I dump the most up to date version of the data to tape? For instance on Friday lets say, I would like to dump my file server (full backup) to tape which happened Tuesday and has had incrementals since then, how do I make sure all of the files after the full get backed up since Veeam doesn't go straight to tape from the server itself?

Thank you again

[Dima P.]

Does rehydration slow down the backup to tape job?

For deduplication appliances (for instance HPE StoreOnce or Dell EMC DataDomain) it does, but I am not aware of any performance issues when it comes to ReFS repository.

is it possible to apply deduplication to the backup to tape job coming from an ReFS volume? (for instance, if the backup job is backing up an entire repository as opposed to one server, because obviously doesn't seem like there would be anything to deduplicate if its just the one server.

Unfortunately that's not possible with tape jobs.

Also due to it's nature, I am guessing Refs doesn't much help with oracle data either? Especially encrypted data? (Kinda talking myself through also as I read and write, so thanks for bearing with me. )

I'd say whenever you encrypt any data do not expect any noticeable savings from deduplication.

If I have a job and it does a weekly full and then incrementals the other days is that two separate jobs?

You can do that via single backup to tape job.

How would I dump the most up to date version of the data to tape?

Tape job always pick up all the backup files you have in the repository, so that should not be a problem. Just make sure job schedule is configured the way you need it.

For instance on Friday lets say, I would like to dump my file server (full backup) to tape which happened Tuesday and has had incrementals since then, how do I make sure all of the files after the full get backed up since Veeam doesn't go straight to tape from the server itself?

Need small clarification here - are you planning to use file to tape (which allows you to protect files directly from the source) or you would go with backup to tape (which is designed to write backup files created by Veeam jobs to tape)?

[MaxDes]

Man, you are on the ball my friend! Thank you!!

file to tape is new? I was under the impression that you could not backup straight from the source to tape, that you needed to backup to disk first?

[MaxDes]

How does ReFS or deduplication work at all if you have Veeam encrypt the backup job or repository?

[Dima P.]

You are welcome!

File to tape is quite old and was designed for file-level only, so no image-based backup here. Backup to tape is more 'sophisticated' job which provides the integration with Veeam jobs in terms of scheduling, synthetic full backup creation, parallel proccing for multiple backup chains to multiple tape drives, but it follows disk to disk to tape approach.

How does ReFS or deduplication work at all if you have Veeam encrypt the backup job or repository?

Need a small clarification here: encryption has no impact on ReFS since block cloning is performed by Veeam B&R, it tracks the identical blocks based on the original content and then instruct ReFS to clone those. Storage level deduplication, however, is affected because data blocks loose it's original pattern as soon they are encrypted and storage device cannot somehow detect it (simply put - two identical blocks are no longer identical as soon as they are encrypted).

Good link to the best practice guide: NTFS or ReFS?

[MaxDes]

Well I guess my question is: How does ReFS work or how can I get the functionality of it, if I encrypt the job using Veeam B&R built in encryption?

[PetrM]

Hi Marc,

It does not matter whether you use encryption or not, please refer to this page for more information about the functionality.

Thanks!