UBNT 2.4ghz throughput
I see people quite often talking about their 10, 15, 30+ meg services that they offer customers, and many say they are doing it using UBNT networks. I personally am not a big fan of UBNT for anything other than a CPE, but I do have one tower I switched to be all M2 equipment as a trial. I have to say that I am baffled as to how anyone could provide those speeds since I've never seen it perform above 15 megs for small peaks, but on average I'm able to get about 10 to 12 megs.
This tower consists of a Rocket M2 with the 13db dual pol omni along with a combination of NSM2's and NB2's for the 9 clients on this AP. I'm running in a 10 mhz channel to avoid overlapping with other devices, so I understand that will obviously cut my throughput in roughly half. Even with signals in the mid -50's, I'm only getting 12 megs of actual throughput right now. The built-in speed test will peak around 25 megs.
Some channels work better than others, but for the most part, I'm using channels with a noise floor in the upper -80's or low -90's. I'm running dual polarity on everything, and Airmax is enabled (disabled seems to drop to about 75% of the speed I get with it enabled) and I'm running the newest 5.5.2 firmware.
The part that I have a hard time understanding is that I have other towers running an XR2 in a 411, standard 802.11g 10 mhz channel, and single polarity which get around 10 megs actual TCP throughput also. I would have expected the 11n combined with dual polarity and a polling protocol would have at least doubled the throughput that I'm seeing.
Anyone have any ideas?
I have a tower that catches noise from a town a mile or two away, and I see results VERY close to what you are describing above, I was very disappointed...
Then an expansion came up later we used 2.4 ghz and it was very rural, I am seeing 20+mbps with a -77signal RocketM2to NanobridgeM2. I was shocked.
The Airview scan on the noisy tower did NOT show much noise, yet it still sucked. I think its entirely based on your noise floor (which is not accurately shown by the UI on the main page).
Switching to Shield Kits from RF armor and throwing the Omni's out helped a ton.
Thanks for the info. I'm not opposed to swapping out to different hardware if that's what it takes. I could do 5ghz for some of the clients, but it is hardly practical to do another AP for 4 or 5 clients. I might try ordering one sector and just seeing how it performs compared to the omni. I suppose I could try the same thing by just temporarily mounting a NSM2 on the tower pointed towards a few of the clients and see what happens.
The performance is fine for now, but I am hoping to offer faster speeds in the near future, and I don't want to do it before I know my network can handle it.
|reply to jcremin |
I was told by a buddy if you plan on doing 10+ meg service you only do about 5 or 6 customers per ap.
Because I am planning on doing 10,12,15 plans.
Those speeds are not an issue for me. I'm talking about going from plans between 768k and 1.5 megs to plans between 1 and 5 megs.
|reply to jcremin |
2.4ghz has the highest noise levels of all your options, RF armour, sectors, shielded Nanobridges, use AirView to look for the quietest spectrum. Use a 20 Mhz channel width if at all possible. I prefer the 5.8 Ghz when line of site allows, but with Airmax on you can get 50 or so clients connected with speeds of 1 - 8 Mbps. I have only gotten maybe 20 2.4 ghz customers on a sector but my 5.8 ghz are hitting 50 clients. I would like to try some of the new rocket titanium radios to compare their capacity.
Your performance is being impacted by your noise floor. I run many sites on ubnt M2 gear and on a clean channel they can do 30 megs, real throughput.
I have one tower in a clean area, rocket M2 + 13 dbi dual pol omni, about 25 customers, most with good signal, around -70 or better. I see a load of around 10 megs in evening, and can run a speedtest.net and get another 20 megs on top of that. I sell up to 7meg service plans and would be comfortable selling 10 or 12 meg plan on this equipment.
In other areas where noise is worse, i do have some sectors that have performance more in line with what you're seeing, 10-15 meg overall AP capacity. The noise floor calculation in the GUI is crap on this equipment, airview is a better tool but I think it doesn't really "see" all types of noise either.
Also, I don't normally see a large difference between built-in speed test and an internet speed test - maybe you have a bottleneck somewhere else in the network?