You've given me a lot to consider. Sent from my iPhone
On Oct 24, 2011, at 9:01 PM, "Tom DeReggi" <[email protected]> wrote: > Can't rely on the noise level reading indicated on a live link's status > screen. (although we tend to put a lot of weight on the noise shown using > the Spectrum Scanner over a period of time, paying attention to the peak > value (blue line) ). > But even then, the results are shown as a single reading, and not > individually for each polarity. What happens if one polarity is low noise and > the other polarity has super high noise? > > For a noise free 2 mile ubiquiti link at 20Mhz, dual pol, we can actually get > Iperf to push 70mbps. Although testing with devices/laptops attached tp ends > of the radios. So if you can only get 40mbps, something is wrong with the > testing method or the link. > > Heck, a 10mhz channel can push 35-40mbps with Iperf, at top modulation.. > You could continue further by dropping down to 10Mhz, and seeing the rate of > change. Also at 20Mhz, you have the flexibilty to change channels to make > sure you aren't centered on a bad channel. > > What speed does the UBNT embedded speed test show? Note UBNT test is also > based on TCP. On average our testing usally will see very simlar speed > results comparing the UBNT built in test to Iperf. If UBNT test is also slow, > you definately have a link issue. If UBNT test is much faster, then scrutinze > your testing devices. Or play with Iperf to make sure its not the Iperf > setting s you are using requiring tweaking. Also, if using Iperf, try > parallel streams, to see if you get a higher agregate throughput. That will > tell you if it is a real capacity limit, or a TCP slow down issue due to > quality. It should be noted that the UBNT Rocket using it's embeeded test > tool, can successfully perform a full speed test. I've tested up to 85mbps. > So, to reduce variables, to trouble shoot a link problem, rely on the > embedded test tool until such time that the embeeded tool can reach full > speed. Once that is accomplished, you can scrutinize link quality further > with Iperf. The only need for Iperf is to test multi-hop to a non-UBNT end > point, or parallel streams, to gather more data. > > We had an issue with Bullets not to long ago, where 5.3.3 firmware was super > slow, but switching down to a 5.2 version solve the performance problem. > But that is not a Rocket. We've used later firmwares for Rockets fine, > usually. > > When you push traffic does your modulation change frequently? or stay steady? > Generally its good practice to leave radio set to a max mode equivllent to > what it will stay at on average without downshifting. > > if your link generally stays steady on a mode/modulation, and you are not > getting throughput appropriate for that modulation, using the UBNT inbedded > tests, its usually a timing thing, preventing the link from functioning > optimally. Manually stting the ACK time is one thing that is within your > control to play with. > When in doubt, if nothing makes sense, and you cant solve it, try changing > firmware. Not only upgrade but downgrade. This has fixed our performance > issues numerous times, even though a clear answer of why, was not determined. > > What antennas are you using? Are they UBNT? If not, do they have a high > enough port-to-port isolation? > > You also cant rule out multipath self interference, as with that you wont see > noise if you aren't transmitting. You can try narrowing antenna beamwidth, if > you aren't already. > > Tom DeReggi > RapidDSL & Wireless, Inc > IntAirNet- Fixed Wireless Broadband > > > ----- Original Message ----- > From: Patrick D. Nix, Jr > To: WISPA General List > Sent: Monday, October 24, 2011 5:52 PM > Subject: Re: [WISPA] UBNT Rocket M5 Throughput > > Ok tried 20mhz, throughput drops about 10mbps to about 30mbps. My signal is > at least 20db better than noise (signal -59 noise -85+). ACK is set auto > with distance of 2 miles (actual distance is approx 1.5mi). cable not an > issue. any thoughts? > > > From: [email protected] [mailto:[email protected]] On > Behalf Of Tom DeReggi > Sent: Monday, October 24, 2011 3:27 PM > To: WISPA General List > Subject: Re: [WISPA] UBNT Rocket M5 Throughput > > You probably either have 1) noisy channels, 2) incorrect ACK distances, 3) > not optimal firmware version, 4) Or some third party factor effecting > testing, such as testing devices that cant generate that much traffic or > buffer sizes of routers. > > The first thing to try is switch down to 20mhz channel and see if the speed > tests stays the same or higher, or if it drops proportionally. I bet the > 20Mhz channel will perform better. > Dont rely on CCQ on its own. Its one indicator, but does not mean you have a > clean channel for sure. Also remember, the Eth port is limited to 100mb, and > if there is cable quality issues such as due to distance, it could autoadjust > to half duplex. Test laptop to PC, isolating RF path, just for grins. It is > very rare to find 40Mhz of clean spectrum for Dual polarity, and even the > slightest packet loss and delay can drastically reduce TCP throughput. > > Tom DeReggi > RapidDSL & Wireless, Inc > IntAirNet- Fixed Wireless Broadband > > > ----- Original Message ----- > From: Patrick D. Nix, Jr > To: WISPA General List > Sent: Monday, October 24, 2011 3:50 PM > Subject: Re: [WISPA] UBNT Rocket M5 Throughput > > Problem is when I turn airmax on the speed really goes in the toilet. Best I > can get is about 20mbps. According to ubiquiti airmax needs to be off up to > 15km links, anything over that airmax needs to be on. Do you have good > success with short airmax ptp links? > > > From: [email protected] [mailto:[email protected]] On > Behalf Of Travis Johnson > Sent: Monday, October 24, 2011 2:45 PM > To: WISPA General List > Subject: Re: [WISPA] UBNT Rocket M5 Throughput > > Turn AirMax on. > > Travis > > On 10/24/2011 1:36 PM, Patrick D. Nix, Jr wrote: > > > What is the best real TCP throughput up/down anyone is getting on a PtP ubnt > connection? We have two rocket M5 approx 1.5 mi, CCQ 97-98%, 40mhz channel > width, airmax off. > > Displayed TX/RX rate is 270/270. Real TCP throughput via iperf radio to > radio is 40-45mbps. > > Are there some config changes needed perhaps? > > > > > > -------------------------------------------------------------------------------- > WISPA Wants You! Join today! > http://signup.wispa.org/ > -------------------------------------------------------------------------------- > > WISPA Wireless List: [email protected] > > Subscribe/Unsubscribe: > http://lists.wispa.org/mailman/listinfo/wireless > > Archives: http://lists.wispa.org/pipermail/wireless/ > > > -------------------------------------------------------------------------------- > WISPA Wants You! Join today! > http://signup.wispa.org/ > -------------------------------------------------------------------------------- > > WISPA Wireless List: [email protected] > > Subscribe/Unsubscribe: > http://lists.wispa.org/mailman/listinfo/wireless > > Archives: http://lists.wispa.org/pipermail/wireless/ > > > -------------------------------------------------------------------------------- > WISPA Wants You! Join today! > http://signup.wispa.org/ > -------------------------------------------------------------------------------- > > WISPA Wireless List: [email protected] > > Subscribe/Unsubscribe: > http://lists.wispa.org/mailman/listinfo/wireless > > Archives: http://lists.wispa.org/pipermail/wireless/ > > > -------------------------------------------------------------------------------- > WISPA Wants You! Join today! > http://signup.wispa.org/ > -------------------------------------------------------------------------------- > > WISPA Wireless List: [email protected] > > Subscribe/Unsubscribe: > http://lists.wispa.org/mailman/listinfo/wireless > > Archives: http://lists.wispa.org/pipermail/wireless/
-------------------------------------------------------------------------------- WISPA Wants You! Join today! http://signup.wispa.org/ -------------------------------------------------------------------------------- WISPA Wireless List: [email protected] Subscribe/Unsubscribe: http://lists.wispa.org/mailman/listinfo/wireless Archives: http://lists.wispa.org/pipermail/wireless/
