Performance of B&R with RDP Server VM

[joachimeberhard]

Hello, I am using Veeam for about 2 months now.

The one thing I noticed is that our relatively small RDP-Server with 25 CALs takes significantly more time to backup or replicate than any other VM including large Fileserver, Exchange, ERP etc. even though it is a very small VM compared to others.

During night hours, when Veeam jobs are running, there is usually no load at all on the Servers or maybe 1 user.

In my replication jobs, most VMS finish between 3-8 minutes, but my RDP is usually slowest with about 20 minutes, which results in that total job length.

Maybe this is normal for several reasons, I just wanted to share my observations.

Edit: It is a Windows 2019 only environment.

Thank you

[HannesK]

Hello,
what are the values for "data read" and "data transferred" for the different servers? I assume that the change rate on the terminal server is just higher than on the other servers. More data = longer time. Or maybe change block tracking is broken for whatever reason (in the job logs there must be a [CBT] next to the disks)

Best regards,
Hannes

[joachimeberhard]

Hm, CBT and transfer rate looks not unusual compared to other VMs.
The amount of data transmitted is even very very low compared to any other VM, usually only 4GB.

Edit: So it says Disk 1 (255GB) 4,1GB read at 154MB/ss (CBT)
I do have VMs that go up to 800MB/s but there are several with only 100ish MB/s, but they finish in minutes.


From what I see in the logs, what takes so long is the creation of the VM recovery checkpoint with app aware processing. (around 15 minutes)

Maybe there is the issue?

[PetrM]

Hi Joachim,

Looks like your version about checkpoint creation is correct as long as it takes 15 minutes out of 20 minutes of overall job duration. However, Veeam does not create a checkpoint but just sends a request to the hypervisor so I'm not sure that there is something that can be done on Veeam side. One more question by the way: does this issue affect your backup strategy or you're just curious about possible reasons?

Thanks!

[joachimeberhard]

I was curious why this particular VM takes longer than any other.
I suppose it is a HyperV thing then.
It is just interesting to see why a RDP server even on low end usage is that much slower than anything else from a 4TB file server to several larger DBs etc.

It doesnt affect my current strategy, because I never had as many great ways to ensure data safety until I bought Veeam anyways
It is just, I haven't had as much fun working and experiencing a new software in like, a very long time.

I was just curious, if I could get replication speed down to a level where I could run it several times a day, but it is not my current goal.
But the things Veeam has tought me made me clear all my servers and arrays and set up a much more robust HyperV environment than I had before...

[joachimeberhard]

I wanted to give some update on this case.
Last month we had the issue that checkpoint creation exceeded 25 minutes in which case the backup aborted.

I found the soulution to be clearing user profiles manually including temp folders.
This reduced the HyperV checkpoint creation time back to 15 minutes and the backup succeeded again.

So there definitely is a performance issue related to HyperV checkpoints on servers with "many" (actually its only 25) user accounts on a RDP server with lots of data in them.

Just wanted to share this observation, maybe it will help someone else.

[PetrM]

Hi Joachim,

Very helpful information for all future readers of this topic, much appreciated!

Thanks!