Physical proxy, Server 2012 R2 or Server 2016?

[HendersonD]

Just received a brand new HPE server that will be a Veeam physical proxy server for SAN based backups
2 CPUs, 16 cores each, 32 cores total, 2.60GHz, 64GB RAM, 10Gbe connectivity
On our current 6 year old proxy server we are running Server 2012 R2
Should we run the same thing on our new box or is there an advantage in running Server 2016?
We are backing up/replicating 40 VMs all running under ESXi 6.5U1

[PTide]

Hi,

If old 2012R2 server works fine and you don't experience shortage of resources, then I don't see any benefits of moving the proxy to the new server. On the other hand, I would recommend to setup a second proxy server on HPE beforehand, to use it as a spare in case the old one goes down unexpectedly. Because, you know, 6 years is 6 years.

Thank you

[HendersonD]

The new proxy server has much better resources than my old one. I will be moving to the new proxy server but may keep the old one around in case of an emergency. I was not sure if Server 2016 had any new technologies built into it that a Veeam proxy would take advantage of.

[PTide]

Windows 2016 is indeed beneficial for repository placement because if its advanced features like ReFS support.

Thanks

[HendersonD]

I have a Nimble all flash array in my data center and a Nimble hybrid array in my DR site. Right now I have a volume called VeeamBackup on the Nimble array in my DR center. My current physical proxy server has this volume mounted as an E: drive. During backup this is where data is written. This E: drive is NTFS

I am foggy on exactly how ReFS fits into this. If I run Server 2016 and mount a volume from my Nimble array can this volume be a ReFS volume? Would this provide some type of benefit doing it this way?

[PTide]

If I run Server 2016 and mount a volume from my Nimble array can this volume be a ReFS volume

Sure, you can place ReFS filesystem atop of LUNs provisioned from Nimble, however there is no safe and reliable way of conversion on the fly. You'll need to create new LUN and format it as ReFS, and move your existing data there. Please take a look at this whitepaper (page 4) to learn about limitations before doing so.

Would this provide some type of benefit doing it this way?

You'll get space savings (especially on synthetic and GFS fulls) and faster transform operations.

Here is the existing thread on usage of ReFS with Nimble, worth checking.

Thanks

[HendersonD]

We have a backup job that runs as forever forward incremental with the target as the volume that comes from the Nimble. We keep 30 days of backup and never do a Synthetic Full. We have another backup job that backs up the same VMs to an off campus Exagrid box at our ISPs location. We keep 10 days of backup on the Exagird and weekly do a Synthetic Full.

I am thinking with our current setup ReFS would not be of benefit. Am I correct? Even if there is not a benefit now, I may go the Server 2016 (ReFS) route now in case we switch things around

[PTide]

In case of forever-forward you'll get faster merges. If you are satisfied with the merges speed that you have now then here is no reason to jump on ReFS right now, indeed. Should you decide to move, make sure to check this thread.

Thank you