VB&R Backup to iSCSI Target

[bsmith-mmo]

My environment consists of 2 locations, HQ and Distribution. Each site has a set of servers with the primary infrastructure located at the HQ site, including the Veeam server, and each site has 2 proxies. Each site has a Qnap that has a 25TB CIFS share carved out. I have backup jobs to backup each site targeting to it's local Qnap, then a backup copy job with GFS targeting to the remote site Qnap for each location. This works, but I'm wondering if I can improve it by converting my backup repositories from CIFS to an iSCSI target with REFS.

For HQ, I'm not sure if I should configure the iSCSI target on the Veeam server and each of the proxies, or just the Veeam server. For the Distribution site since it does not have the Veeam server local, should I configure the iSCSI target on the Veeam server or on one or both proxies?

HQ:
Veeam
2 Proxy
HQ-Qnap

Backup job -> HQ-Qnap via CIFS
Backup copy job: HQ-Qnap -> Dist-Qnap via CIFS


Distribution:
2 Proxy
Dist-Qnap

Backup job -> Dist-Qnap via CIFS
Backup copy job: Dist-Qnap -> HQ-Qnap via CIFS

[Egor Yakovlev]

Hi Brian,
I assume your current Repository(CIFS) is determined as a bottleneck component in job session statistics window? Because if it is not, and bottleneck is a network, for example, even if you change to full-SDD FC connected SAN instead of CIFS, nothing will change performance wise.
As for the iSCSI mount, use local-to-site servers, aka mount on Veeam server in HQ, and mount on one of the proxies on Dist. You don't need to mount same QNAP box on both proxies, cos we will treat them as 2 different repositories in this case.
/Cheers!

[bsmith-mmo]

Most of the time the bottleneck is the target since it's HDD. Sometimes the network, rarely, if ever, the source since it's SSD.

I'm assuming it would be best to leave the design alone since there really wouldn't be any performance increase due to the bottleneck generally being the target?

[Egor Yakovlev]

Well, design is still a good topic, especially in a multi-site case.
Check if you have specified correct Gateway\Mount server for your Dist QNap Repository settings, cos if you forgot to do so, your Dist backup traffic goes Dist->HQ Veeam->Dist Qnap.

[bsmith-mmo]

I do have the Dist-Qnap repository gateway\mount server assigned to one of the local dist proxies.

[bsmith-mmo]

I ended up attaching the iSCSI to VMware and then create a datastore on the Qnap iSCSI target. From there I created a new virtual disk that I attached to the Veeam server, formatted it with ReFS, then disconnected the disk and attached it to a local proxy.

With this design I have seen a massive reduction in backup times with free synthetic fulls as a bonus. Previously, a full backup of my 2.1 TB file server would take between 9 hours and 3 days, now it will complete a full backup of the 2.3 TB file server in 3.5 hrs. Going from a single 1Gb to a lagged 10Gb connection has been a huge benefit as well.

[Egor Yakovlev]

Great news and thanks for sharing your experience.
/Cheers!