We are trying to setup a simple bonding with two wireless bridges (4xSXT + 2xRB493G). The result was a poor performance. We though we were doing something wrong with the wifi setup, so we decided to do it with a couple of UTP links. The bonding performance was even lower that the simple link.
The bonding setup seems to be very simple. Is there something special we have to consider?
We tested the performance with two computers running btest. Each PC was connected to a RB493.
A bit of help will be appreciated. We are planning our OSPF backbone and the speed of links is one of our main goals.
@cbrown
We are triying 2xUTP connection. When we get good results, we will jump to wireless. In our previous tests, we built the wireless bridge using MPLS tunnels. The CCQ was good in each single link. Both links where in very separated channels.
@Inox
This was our first approach. Two transparent bridges with 4xNanostationM5. Single link was faster than bonding.
Thinking about compatibility issues, we built the bridge with mikrotik in wds, and again with a mpls tunnel (http://wiki.mikrotik.com/wiki/Transparently_Bridge_two_Networks_using_MPLS).
Allways the single link was fastest than bonding.
So we decided eliminate the wifi and concentrate in the bonding via cable, with rr-balancing mode.
as result you will have full duplex wireless link (one pair of radios will receive and other send traffic. but you require 2 radios in one device. IMHO that is better for wireless links than bonding, that were designed for full duplex links that wireless is not.
Some of our links runs nstream dual. But in this case, we are not looking for full duplex. We talk about typical internet traffic, with a great asymmetry. We need the best speed for download.