I have been experimenting with different ways to make a transparent bridge across two 5Ghz-N-only units.
Hardware: 2 x RB600 with UBNT SR71-A
Test tool: 2 x RB1000 as “RB1000 - RB600 - RB600 - RB1000” - all bridged together in one subnet.
All tests have been made indoor with 2x2 Mimo config. Signal levels adjusted to approx. -50 dB, which gives max airspeed at both interfaces.
The best results I have been able to get is when I configure the radios as plain AP/bridge. The L2-bridge is created as eoip. In this case I get 50+ Mbps TCP fdx (as 50/50, 40/60, 20/80, 0/100 or any other). All 4 units are running at 100 % cpu load in this situation, and are probably the limiting factor.
Any other configuration (with/without nstreme, wds, etc) I have tested, yields substantially less.
Compared to our normal rates on bridges, these results are not bad. However, 300 Mbps should bring more real life speed ???
Any comments ???
PS - please don’t start a discussion regarding the test method. A tcp-test correlates very well with real production results.
Sad to see nobody gives you any feedback yet. Like me I’m eager to see some answers.
What struck me in your setup is that all rb’s are running 100% cpu?.
I would expect the rb1000 would have no problem producing your data througput.
If I use a rb1000 behind a rb433AH and get 35Mb over the link the rb1000 shows no more then 3-5% cpu.
this in a routed setup but I wouldn’t expect bridging or routing to make such a difference in just producing data to send…?
I have just been playing with some of this stuff this evening. I had this config:
Bridge in wireless setting
create new eoip interface from wan to wan on other radio ip
create new bridge, add eth1 and eoip tunnel
next radio
station in wireless setting
create new eoip interface from wan to wan on other radio ip
create new bridge, add eth1 and eoip tunnel
With n-only and no additional control channel (20Mhz) I was able to get 50MB tcp either way at one time.
With additional control channel above (40Mhz) I was able to achieve almost 70MB tcp.
These are with a RB411. I used a RB433AH and my readings were higher.
WDS is not an option with N mode. I tried to see what the limitation was and it just plain didn’t work well at all and was inconsistent. EoIP seems like a viable alternative that is easy to implement.