I’m trying here to set up some short distance p2p link using a pair of SXT’s. The distance is about 1km, clear LoS, signals are ~-51/-53 all the time. Unfortunately I struggle achieving real life speeds higher than 38-41mbps. I would take as much as possible, at least 70mbps in TCP is ok for me. I’ve read in this forum, that people are experiencing similar problems with bandwidth ceiling at around 30mbps. So I would like to make a list of questions to be answered by users that could share their own experience or knowledge how to make simple p2p link with as higher bandwidth as possible using 11n, nstreme/nv2 and whatever else needed.. Hope this also serves as a guideline for newcomers.
So here are few questions:
First of all, is SXT (regarding CPU speed) enough for up to 100mbps of traffic (eth limits it to that number anyway)?
What kind of link setup you should choose to achieve high speeds: routing, bridging (wds or mpls)?
Do you really need two chains and/or extended channel for 100mbps?
How much difference is there while testing such setup in laboratory environment and real life scenario?
Do you need to tamper with Data rate selection?
I’ve tried every setup example provided here in this forum, but no luck really. I am aware that 11n is much more to understand and tweak, but to get no more than 40mbps whatever you do in lab environment is a little bit weird. Testing was done by two x86 machines on ethernet side of SXT’s using web server and Speedtest Mini. I don’t really want to stick with 11a Turbo mode forever, even if it provides about 55mbps.
try nv2 and 40mhz below or above, mpls is more efficent, put data rate configurated an select mcs on hight data rate. on advance select hwretrieves and change to 15.
Currently I’m trying to find a good working setup in lab. environment. But I do have SXT’s already mounted in-place of my actual link, and I did choose the best and cleanest channel for it. Signals are all at around -51, give or take 3-4dB. CCQ are >90% both ways. And the test environment is completely radio/noise free.. I can get signals whatever I want, CCQ is 100/100% all the time. I’m testing it in a 30m long room which is located in the basement. Maybe that might be my problem?
I already mentioned how the bandwidth speedtest was done in my first post. Two x86 machines, one with web server serving speedtest mini flash app and other simply doing that test using a web browser. No bottle necks here, if I connect them directly it throws 98mbps.. Good old R5H’s in turbo mode showed ~55mbps in that same test.
JorgeAmaral, what do you use? Routing or bridging?
VPLS tunnel between the SXT. Dont forget to adjust MTU of the MPLS to be larger that 1526, or you will suffer from fragmentation.
Normally i set my wlan mtu to 1534, and mpls mtu to 1534 too.
When you are doing a bw test, see the amount of pps on the ethernet interface and on the wlan interface, they should be equal. If you have much more packets on the wlan, then you have fragmentation issues and this will slow down your goodput.