Optimising Repositories for Backup Copy

[readie]

I have some very large backup copy jobs (6TB) and am trying to optimise the arrangement of Repositories.
My backups are sitting on an iSCSI SAN (SAN1) and the copy is going to another iSCSI (SAN2).
My repositories are Windows 2012 servers using iSCSI to connect to the SANs.
My question - is it better to have both SAN1 and SAN2 connected to the same Repository server, and have it do ALL the work of the Backup Copy job, or connect SAN1 to RepositoryA and SAN2 to RepositoryB so there are two servers sharing the load (rather like source proxy and target proxy).
(Am I right that Backup Copy will not use Backup Proxies, and that the data transfer will not have to go via the Backup Server itself? Or does the data transfer go via my Backup Server?)
The repositories are VMs, and if I force RepA and RepB to be on the same ESXi host, my understanding is that all network traffic between them will be at the Virtual Switch level, so will never get out onto our physical network - though this is 10GbE if required.
Thanks for any hints and tips - before I start 'trial and error'!
Bob

[foggy]

Backup copy job traffic is transferred directly between repository servers (so if those are dedicated servers, then neither proxies, nor backup server are involved). What is currently the bottleneck for the backup copy jobs? Unless it is network, combining both data mover agents on the same server is out of help.

[readie]

Hadn't thought of that - Source is the bottleneck at the moment. At the moment both iSCSI are connected to single Repository server, and Source 98%, Proxy 34%, Network 2%, Target 6% (but I thought no Proxy was used for Backup Copy?).
So probably best to share the load between 2 Repositories? Is there much processing going on, and if so mostly on the source or the target?

[foggy]

"Proxy" in case of direct backup copy job is Veeam data mover agent installed on the source backup repository. According to these stats, your source SAN is the bottleneck (probably is busy with some other tasks), so I don't think adding second server helps. Processing load depends on the compression settings mostly.