Rebuilding a SAN for backups - iSCSI or CIFS/NFS?

[chiller15]

I manage a QNAP TVS-1271U-RP SAN in my work and use VEEAM as our backup software, which is connected via 10GbE to our production storage (HP MSA). We've maxed out our storage capacity and temporarily migrated the data to a spare datastore. We have replaced the existing 9x 3TB drives (in RAID5 with a CIFS share) with 8x 8TB drives in RAID6 (for extra redundency). My question is; Do I use iSCSI or a standard file share like CIFS/NFS?

My new RAID6 array is currently synchronising, so I have to wait until tomorrow when that completes to proceed with the configuration. I have a basic understanding of iSCSI and that it performs better than a standard file share, though I'm not sure if there would be any downsides.

[Windows7ge]

I'm not the best person to be answering your question so take what I say with a grain of salt but from my experience with both protocols they have pros & cons. iSCSI specifically I have experienced great performance with but at the same time I've had it mysteriously drop to a near halt and crawl painfully to the finish line. That could just be a hardware/software config issue on my part though.

 

SMB hasn't performed as well (peaks) but it's also never tanked as hard as iSCSI. At the same time though I have had SMB just give out mid transfer and had to restart when it left off. (Software like FreeFileSync or RSYNC make finding where it left off really easy)

 

If it was a lot of tiny files I'd go as far as to recommend SSH/SFTP. Although I have had sessions cut out it has been by far the most reliable protocol. Downside is it is by far the slowest. Not great for massive files on a 10Gbit network.

 

So long as a copy of the data will exist and you're not Cut/Paste-ing I'd say it doesn't really matter which method you use (if you can't connect all the drives to one server and copy RAID A's data to RAID B directly over SAS/SATA). Personally I'd use SMB and if it stops half-way though use FreeFileSync or RSYNC to find where it left off and continue the transfer.

[leadeater]

If you do switch over to iSCSI you can go with the current Veeam recommendation of using ReFS instead of NTFS, have a look at Veeams documentation on the benefits of using ReFS.

[chiller15]

5 hours ago, Windows7ge said:

I'm not the best person to be answering your question so take what I say with a grain of salt but from my experience with both protocols they have pros & cons. iSCSI specifically I have experienced great performance with but at the same time I've had it mysteriously drop to a near halt and crawl painfully to the finish line. That could just be a hardware/software config issue on my part though.

 

SMB hasn't performed as well (peaks) but it's also never tanked as hard as iSCSI. At the same time though I have had SMB just give out mid transfer and had to restart when it left off. (Software like FreeFileSync or RSYNC make finding where it left off really easy)

 

If it was a lot of tiny files I'd go as far as to recommend SSH/SFTP. Although I have had sessions cut out it has been by far the most reliable protocol. Downside is it is by far the slowest. Not great for massive files on a 10Gbit network.

 

So long as a copy of the data will exist and you're not Cut/Paste-ing I'd say it doesn't really matter which method you use (if you can't connect all the drives to one server and copy RAID A's data to RAID B directly over SAS/SATA). Personally I'd use SMB and if it stops half-way though use FreeFileSync or RSYNC to find where it left off and continue the transfer.

The storage will be used as a backup repository for VEEAM, which will create and manage large backup files itself. Once I've copied the paused jobs to the new storage, VEEAM will continue to manage them. I've read that VEEAM works best with iSCSI repositories, but I'm unsure as to why compared to SMB.

2 hours ago, leadeater said:

If you do switch over to iSCSI you can go with the current Veeam recommendation of using ReFS instead of NTFS, have a look at Veeams documentation on the benefits of using ReFS.

That's what I was thinking, but I also read that ReFS can have issues and bugs once the storage is more than 50% full, even when on Server 2019 (which the VEEAM server is). Is that the case?

[Windows7ge]

8 minutes ago, chiller15 said:

I've read that VEEAM works best with iSCSI repositories, but I'm unsure as to why compared to SMB.

Unfortunately I can't verify that for you. I have more experience with Linux/*NIX than I do anything Windows Server related.

[leadeater]

29 minutes ago, chiller15 said:

That's what I was thinking, but I also read that ReFS can have issues and bugs once the storage is more than 50% full, even when on Server 2019 (which the VEEAM server is). Is that the case?

Not that I know of but it's something you can just ask Veeam support.

[chiller15]

3 minutes ago, leadeater said:

Not that I know of but it's something you can just ask Veeam support.

Unfortunately we don't have a valid support contract with VEEAM, so I'll post on their forum in the morning and see what responses I get.