We don’t believe there is any software bug here, and it is important to note that the speed tests were performed while the link was heavily utilized. PHY rates were 1300 Mbps, and existing throughput was 766 Mbps aggregate at the time. These details were left out earlier in the thread.
Auto TDMA provides half the latency of a fixed TDMA with 8ms window, and as we know, TCP performs better with lower latency. Switching to Auto TDMA allocates more capacity in the direction where it is needed, further reducing latency for all clients, and improving TCP speed test results. Mimosa always recommends running speed tests on links in isolation to identify the source of any bottlenecks. This usually entails running a standard test such as iPerf across the link with nothing in between. We have nothing against speedtest.net, and regularly see speed test results that approach link capacity when the link throughput is not near capacity. With fixed TDMA time slots, the MAC rate and performance are consistent, but latency is higher. Here is an example using the settings we saw on that specific link: Tx: 1300 PHY * 90% Efficiency with 8ms * 75% traffic split = 877.5 Mbps MAC rate (658 with TCP overhead) Rx: 1300 PHY * 90% Efficiency with 8ms * 25% traffic split = 292.5 Mbps MAC rate (219 with TCP overhead) Latency = 8ms TDMA window * 2.5 = 20 ms With Auto TDMA, the MAC rate shown on the Dashboard does not represent the full Auto TDMA capacity. It shows a 75% traffic split in both directions in all conditions, while Auto TDMA is capable of traffic splits beyond that. Thanks, Chris for pointing that out, and sorry for the misunderstanding. Here is the calculation that the Dashboard shows. Tx: 1300 PHY * 80% Efficiency * 75% traffic split = 780 Mbps MAC rate Rx: 1300 PHY * 80% Efficiency * 75% traffic split = 780 Mbps MAC rate Latency = 1-10ms depending on load The Tx MAC rate appears fixed and lower in Auto TDMA, but is isn’t necessarily. Instead, the MAC rate could be as much as the calculation below since Auto TDMA adapts the split to traffic demand: Tx: 1300 PHY * 80% Efficiency * 90% traffic split (auto) = 936 Mbps MAC rate Chris Trout Mimosa Networks, Inc. From: Af <af-boun...@afmug.com> on behalf of Gino Villarini <g...@aeronetpr.com> Reply-To: "af@afmug.com" <af@afmug.com> Date: Wednesday, January 25, 2017 at 10:49 AM To: "af@afmug.com" <af@afmug.com> Subject: Re: [AFMUG] B11, TDMA, and TCP Ouch, I’ve been slapped! Lol! Thanks Tim for reminding us stuff we forget with age From: Af <af-boun...@afmug.com<mailto:af-boun...@afmug.com>> on behalf of Tim Hardy <tha...@comsearch.com<mailto:tha...@comsearch.com>> Reply-To: "af@afmug.com<mailto:af@afmug.com>" <af@afmug.com<mailto:af@afmug.com>> Date: Wednesday, January 25, 2017 at 2:46 PM To: "af@afmug.com<mailto:af@afmug.com>" <af@afmug.com<mailto:af@afmug.com>> Subject: Re: [AFMUG] B11, TDMA, and TCP “Or mis-align your B11 link” Strictly illegal! From: Af [mailto:af-boun...@afmug.com] On Behalf Of Gino Villarini Sent: Wednesday, January 25, 2017 1:44 PM To: af@afmug.com<mailto:af@afmug.com> Subject: Re: [AFMUG] B11, TDMA, and TCP AF24Š Or mis-align your B11 link On 1/25/17, 2:36 PM, "Af on behalf of Daniel White" <af-boun...@afmug.com<mailto:af-boun...@afmug.com> Gino Villarini President Metro Office Park #18 Suite 304 Guaynabo, Puerto Rico 00968 [cid:image001.png@01D27705.9B00FA00] on behalf of afmu...@gmail.com<mailto:afmu...@gmail.com>> wrote: >Not with a B11. > >If the equipment is at the minimum TX power and that high of a SNR is >going to cause problems... then it¹s the wrong gear for that path. > >Siklu or SIAE 80GHz would be my first bet for a 2.2km link. > >Daniel White >Managing Director Hardware Distribution Sales >ConVergence Technologies >Cell: +1 (303) 746-3590 >dwh...@converge-tech.com<mailto:dwh...@converge-tech.com> > Gino Villarini President Metro Office Park #18 Suite 304 Guaynabo, Puerto Rico 00968 [cid:image001.png@01D27705.9B00FA00] >> -----Original Message----- >> From: Af [mailto:af-boun...@afmug.com] On Behalf Of Seth Mattinen >> Sent: Wednesday, January 25, 2017 9:36 AM >> To: af@afmug.com<mailto:af@afmug.com> >> Subject: Re: [AFMUG] B11, TDMA, and TCP >> >> On 1/25/17 09:30, Faisal Imtiaz wrote: >> > Thanks for pointing that out ! >> > >> > So any thoughts on how to 'attenuate' the signal ? >> >> >> I would guess waveguide and attenuators, but I don't have any personal >> experience with a B11. >> >> ~Seth > > >--- >This email has been checked for viruses by Avast antivirus software. >https://www.avast.com/antivirus >
