Veeam Replication Server Network

[bproteau]

We are using VMs as the Veeam Backup and Replication server for a one time Site to Site migration (Dallas-Phoenix).

(1) Is the replication going through the Veeam server or is it Source Host - Destination Host communciation.
(2) Related to that question, I'm trying to determine the best network to put the source side servers on. If repliction is throught the Veeam server, I planned on putting Veeam server(s) in the same VLAN as the destination Hosts to avoid and firewall between sites.

Environment:

- Source ESX Hosts and vCenter (VLAN 10)
- Destination ESX Host and vCenter (VLAN 20)
- VLAN 20 exists in both source and destination locations
- 10GbE connection between sites (35-40ms latency)

[Gostev]

In v5, replication always goes through the Veeam server, because the Veeam server is what pulls the source data.

[bproteau]

Thanks...that was what I assumed so, it probably makes sense for me to put the Veeam Servers on the same network as the destination Hosts so I can avoid any firewall complications or limitations on the long distance leg (Dallas Veeam Server - Phoenix Hosts). Locally, (Dallas Hosts - Dallas Veeam Server) data will be pulled through a firewall but, in my mind, this seems a better configuration than the reverse.

I'd appreciate any other comments with respect to this network configuration design pros or cons.

[Gostev]

The best recommendation here would be take to scrap idea of deploying v5 replication, and go ahead with v6 that is just around the corner. If you start deploying now, by the time you have everything planned you would have to get back to the drawing board. v6 has vastly enhanced replication, but its architecture is quite different.

We have been running private beta with a few selected customers, and the feedback has been overwhelmingly positive. Think beta customers like one of the largest European banks, testing replication between very geographically dispersed sites (as far away as you can have them in Europe, pretty much).

[bproteau]

I've been hearing "just aroud the corner" for some time. I'd like to use v6 but, are migration timeline iss compressed and without a specific release date, we need to start replications almost immediately.

[Gostev]

I've been hearing "just aroud the corner" for some time.

Well, considering this was the very first time I personally said this, there's hope that it really is there!

Other than that, your plan looks good. The fact that you have "fat" ESX hosts for targets makes everything really easy.
Meaning, this lets you be flexible regardring Veeam server placement.

[bproteau]

I should have been more specific. I do have ESXi 4.1 U1 and vCenter 4.1 U1 on both sides.

[Gostev]

In that case, you need to place your Veeam Backup server close to the target hosts in order to get reasonable performance. Check out the adjacent topic (replication and ESXi) for more details on replicating to ESXi with v5.

[bproteau]

Sounds good. Appreciate the assitance. When we talk pull replication, does that just mean the Veeam Hosts are in the destination site? I'm also looking for the technical reasoning why a pull is better. The network between these 2 sites is a bit convoluted at the moment with some destination connectivity being backhauled up the 10GbE link to services in the source.

I'll look for these answers in the thread you referenced. The more I know about precisely why pull is better, the better I'll be able to plan and design the Veeam Replication network topology for our specific environment.

[bproteau]

I had and read that post again. It is a good discussion but, I don't think I read any details on specifically why performace is improved with pull vs. push for ESXi.

In addition, I read that a push is not recommended for ESXi. Will I have issues if I used a push with ESXi or is Veeam just saying that pull is "preferred"?

[Gostev]

No issue, just 3x more traffic (and so 3x worse performance on WAN link). Which in turn directly translates into reliability (as WAN links tend to drop).