Ive recently inherited some aging infrastructure which is struggling to complete backups in any sensible times.
We have a physical Veeam server running 9.5.0.1536 which has 10Gb eth connected to 10Gb switch and 10Gb backbone to our blade clusters. This backs up approx. 200VMs each night over 12 jobs.
This has 10Gb connectivity to the storage NAS.
The VMware estate is still on 5.5 but will be moving to 6.0 shortly to put us back under a supported platform. Due to hardware, its unlikely that we can go to 6.5 without a hardware refresh and the man with the purse strings is cloud indoctrinated (delusions of budget savings), so the refresh is unlikely right now.
The NAS is QNAP TVS-EC2480U-SAS with 24x 4TB disks in a single volume of ~80TB, with 2x internal cache modules used for acceleration.
An internal volume has been created which contains a multitude of data including shared folders used by the business as well as shares for Veeam backup repos.
The good news is its 10Gb with lots of spindles and some cache acceleration.
There are 4x CIFS shares from this volume which are mounted on the Veeam server as separate repositories. "\\nas\Veeam1\" etc..
So back on the Veeam server, Ive updated each of these 4x repos to limit concurrent tasks to 8 each (yes I know this is pretty pointless as the 4x repos are backed by the same storage, but Im looking forward), and have set the repos to per-VM.
Most of the jobs were reverse, but I have updated most of these to forwards with weekly synthetics, which seems to be working well on the smaller jobs.
I have a single backup job of a DFS server (2008r2) around 5.5TB in size which is taking 3 days to complete a full as and when it fails, and typically 22-26 hrs for its nightly backup which regularly fails or overruns. I have tried switching this over to forwards with weekly synthetics, however the run-time on the synthetic is 3-4 days which results in missed backups for a couple of days, so that got switched back to reverse as at least we get some backups with that.
I am thinking about presenting an ISCSI LUN from the same NAS to the server to use as a repo to move away from "QNAP" format filesystem for starters, but if Im going to that effort, I can just build a new Veeam VM running 2016 server and format the iSCSI LUN as REFS to leverage block clone savings from REFS?.
My questions are, will this actually gain me any significant benefit, or does anyone have other recommendations on performance tuning what I have?
Thanks
Owen