Best strategy for syncing multiple LAN nodes across a single bottleneck WAN connection

What is the best strategy to sync multiple nodes between two fast LANs across a single bottleneck slow WAN connection?

Is is faster to only sync a single node across the bottleneck WAN and let that node then share changes across the LAN? Or is it just as fast to sync all nodes across the bottleneck WAN thereby sharing the bottleneck WAN connection, and then having the partial syncs combine on the LAN side? Am unclear if syncthing partial syncs will keep syncing on the LAN side as other nodes on the same LAN update separately across the bottleneck WAN. If unclear will rephrase. Thanks.

In my opinion you will be best off having only one pair of nodes sync across the WAN. Additional nodes will contend for bandwidth. Even if they share the data, there will be metadata and link overhead.

I agree with above, best not to overload the bottleneck. I would use four nodes in total, two on each LAN, connected to each other, for redundancy.

@Corrisen If one uses multiple nodes for redundancy the WAN connection is going to be split between the redundant nodes (not necessarily bad, but the question is whether this slows down the syncing process, and consequently is not best practice.)

The consensus seems to be that:

Node A (LAN 1)

Node B (LAN 1) <--------SLOW WAN--------> Node D (LAN 2)

Node C (LAN 1)

is preferable to

Node A (LAN 1) <--------SLOW WAN--------> Node D (LAN 2)

Node B (LAN 1) <--------SLOW WAN--------> Node D (LAN 2)

Node C (LAN 1) <--------SLOW WAN--------> Node D (LAN 2)

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.