Replication slow? Am I doing something wrong?

[dsegel]

I'm using VB&R v11 to to do continuous replication of 5 servers over a slow (100Mbit) link using WAN Accelerators. The servers are a mix of linux servers and Windows servers, including a SQL Server. The servers range from ~ 50GB to 3TB in size (most being under 200GB).

With Hyper-V replication we can keep up with a 5-minute replication interval even over the slow link - none of the servers on the remote side are ever more than 5 or 10 minutes old. However, if there's a fault and it needs to be re-synched it can take several days.

With VB&R replication, the initial sync happened relatively quickly, but the job takes 1 1/2 hours to complete one round. Is this typical? Is there something I'm doing wrong? I had assumed that using the WAN Accelerators would help with this situation.

Let me know what other info you need to help diagnose this...

[Mildur]

Hi

If HyperV Replication only takes a few minutes, then veeam shouldn‘t take much longer. I would say it‘s unexpected. But It‘s difficult to make an assumption without knowing the entire infrastructure and seeing the actual job logs.

Start with analyzing the bottleneck. Is CBT used for the replication or does the replica job read the entire data again before replicating only changes? What‘s the change rate of your vms?

Could you please open a case with our support to let them analyze the debug logs. Our support engineers can tell you what is happening in your infrastructure. There can be many reasons.

Thanks
Fabian

[david.domask]

Just to confirm, when you talk about WAN Accelerators, you mean the Veeam infrastructure items like this: https://helpcenter.veeam.com/docs/backu ... ml?ver=110

Or you have some WAN Acceleration hardware in your network like Riverbeds?

If it's a Veeam one, keep in mind that by default the Source WAN accelerators only process a single disk at a time; you can work around this with multiple Source WAN Accelerators (Many to One), or just do a direct connection (with the High-Bandwidth mode in v11, 100 Mbit is viable with WAN acceleration, but keep in mind that for incremental changes, the savings may still be negligible). But that would be my initial guess on what adds the additional processing time. You should be able to see this in the job statistics that it's only processing a single disk at a time

If it's some hardware WAN accelerator, then likely indeed there needs to be a bit more research on what's taking the longest time based on the bottleneck, and a Support Case is the best approach.