Recommended link speed for DR replication

[AndyR]

Was wondering, is there a recommended ideal link speed for Veeam Replication between PROD and DR for replication?

Reason for asking is we are setting up a new DC (both prod and DR) and are using Veeam for replication. At present the link speed between prod and DR is a dedicated 2mb MPLS. I originally setup our Prod VMs and then copied the backup from Veeam over to DR over a weekend, then restored them in DR and setup replication to map to replicas. When I then start replication going its saying bottleneck 99% network and the CBT part of the job on reading each virtual disk is pegged at 282 kb/sec which is basically the max speed of the link. Our first replication job has been running for 72hrs now and is 12% through.

I'm certain the link is the problem because the job isn't running faster than the link, its just there hasn't been many changes on these VMs since I transferred the backup so I'm surprised it wants to replicate so much. So any suggestions on what you guys use for links between Prod/DR and also any ideas on if I can speed it up if at all possible?

Thanks,

Andy

[Gostev]

is there a recommended ideal link speed for Veeam Replication between PROD and DR for replication?

Only depends on amount of changes VMs you need to replicate generate, and required replication schedule.

[AndyR]

Thanks. How does it compare the changes anyway? Does it look at local changes since the last replication and then send the combined changes over the link in one go, or does it compare changes realtime during the job and find a change-send it-find next change-sent it etc?

[Gostev]

look at local changes since the last replication and then send the combined changes over the link in one go

This.

[dellock6]

Hi Andy,
replica uses same VADP libraries as backup, so Veeam first takes a snapshot of the VM that needs to be replicated, and that will be the status that will be confronted with the VM at DR site, and synced. Veeam does not do CDP (continuos data protection) as you described.
You can simulate the amount of data it will be transferred by configuring a local replica between two hosts and using wan compression level. You will be able to see in the job log the data amount created daily.

Reading your first post, some doubts:
- did you configured TWO veeam proxies, one at each site? This is mandatory for having good performances
- I'm not sure about your numbers, 2 mbits line means a max speed of 256 KB (kilobytes), 282 is more than this speed but you wrote lowercase b, you mean 282 kilobits? This is really low indeed, if it's true

Luca.

[AndyR]

Thanks Gostev.

Luca - Thanks, makes sense now. Yeah I think I was getting confused with how CDP works so now I understand how Veeam does it better. Yep I have 2 proxies. On the transfer rate I think I might be seeing average transfers at the minute hence the wrong figures, so i went back through the job and looked at the disks which have completed so far and they show 246 KB/s so that seems more like the 2MB link to me.

[dellock6]

Ok, now numbers sounds good. Veeam is correct telling you bottleneck is network, you need now to check how much data you have to transfer daily to do your maths and find out if the duration of the replica is acceptable. If not, the only solution you have is yo upgrade your link, or maybe try a wan accelerator.

Luca.