QCA9880 slow performance adhoc/IBSS mode

Ralph Houthoofdt ralph.houthoofdt at ugent.be
Wed Mar 1 00:21:09 PST 2017


Dear Ben,

Indeed, if I try AP/STA mode I get better bitrates and speeds (this is with
VHT40)
RX: 400.0 Mbit/s, 40MHz, VHT-MCS 9, VHT-NSS 2, Short GI
TX: 300.0 Mbit/s, 40MHz, VHT-MCS 7, VHT-NSS 2, Short GI

I only have 4 antennas installed on my APUs, so I'm only using 2 spatial
streams max.
These speeds seem ok.

It just seems strange that on adhoc I only get either RX or TX to have good
rates.

Thanks for your help!

Met vriendelijke groeten,
Ralph Houthoofdt
Student industriële wetenschappen informatica

-----Original Message-----
From: Ben Greear [mailto:greearb at candelatech.com] 
Sent: dinsdag 28 februari 2017 23:03
To: Ralph Houthoofdt <ralph.houthoofdt at ugent.be>
Cc: ath10k at lists.infradead.org
Subject: Re: QCA9880 slow performance adhoc/IBSS mode

I've pushed 475+Mbps on apu2 over wifi, maybe more.  That was AP mode, most
likely.

IBSS on ath10k cannot do AMSDU, so you loose a bit of performance there, but
otherwise it should be pretty similar to AP/STA mode.  To verify, try
running in AP/STA mode.

Thanks,
Ben

On 02/28/2017 01:48 PM, Ralph Houthoofdt wrote:
> Dear Ben,
>
> The speeds I'm reporting are UDP using these commands for the client:
> Iperf3 -uVc <serverAddress> -b0 -w128k
> Iperf3 -uRVc <serverAddress> -b0 -w128k
>
> I will try relocating the nodes and try different antenna positions 
> and angles.
> As you see it, this problem is purely because of the environment, not 
> because of the configuration or drivers?
>
> I will let you know what the results of relocation are tomorrow. 
> Thanks for your help!
>
> Kind regards,
> Met vriendelijke groeten,
> Ralph Houthoofdt
> Student master industriële wetenschappen: informatica
>
> -----Original Message-----
> From: Ben Greear [mailto:greearb at candelatech.com]
> Sent: dinsdag 28 februari 2017 19:12
> To: Ralph Houthoofdt <ralph.houthoofdt at ugent.be>
> Cc: ath10k at lists.infradead.org
> Subject: Re: QCA9880 slow performance adhoc/IBSS mode
>
> You try UDP yet, or is this all TCP?
>
> Also, with some better antenna positions and/or better other RF 
> environment, you should see NSS of 3, hopefully MCS of 6 or higher.
>
> On 02/28/2017 10:02 AM, Ralph Houthoofdt wrote:
>> Dear Ben,
>>
>> I ran the same test again, this time running the iperf3 server on my 
>> laptop and the client on the node. I could indeed see CPU usage rise 
>> to 40% on my laptop, the node stayed at roughly 10%. The speeds 
>> remained roughly the same. 215Mbit one way, 40Mbit the other.
>>
>> As for the tx/rx rates I get these values:
>> tx bitrate:     234.0 MBit/s VHT-MCS 3 80MHz VHT-NSS 2
>> rx bitrate:     87.8 MBit/s VHT-MCS 2 80MHz VHT-NSS 1
>>
>> I set the 4 antennas facing different directions approximately at 135 
>> degree angles relative to the device.
>
> You try UDP yet, or is this all TCP?
>
> Also, with some better antenna positions and/or better other RF 
> environment, you should see NSS of 3, hopefully MCS of 6 or higher.
>
> The 9880 has 3 antenna, so how did you get to 4?
>
> Thanks,
> Ben
>
>>
>> Thank you for your help so far. I will be leaving home for a few 
>> hours and will be responding to any replies tomorrow morning.
>>
>> Met vriendelijke groeten,
>> Ralph Houthoofdt
>> Student industriële wetenschappen informatica
>>
>> -----Original Message-----
>> From: Ben Greear [mailto:greearb at candelatech.com]
>> Sent: dinsdag 28 februari 2017 18:42
>> To: Ralph Houthoofdt <ralph.houthoofdt at ugent.be>
>> Cc: ath10k at lists.infradead.org
>> Subject: Re: QCA9880 slow performance adhoc/IBSS mode
>>
>> Try UDP to make sure you are not hitting a TCP issue.
>>
>> Try running iperf server on another laptop so that CPU of your nodes 
>> is less important.
>>
>> Check the tx/rx encoding rates to make sure you are getting good RF 
>> behaviour.  'iw ... station dump'
>>
>> Ensure antennas are not pointing all in the same direction so that 
>> you can get some diversity working.
>>
>> Thanks,
>> Ben
>>
>> On 02/28/2017 09:36 AM, Ralph Houthoofdt wrote:
>>> Dear Ben,
>>>
>>> Thank you for your quick reply.
>>>
>>> My mesh is running on channel 36  using VHT80 and a TX power of 20 
>>> dbm, the maximum allowed according to my regulatory settings from iw 
>>> reg get. The hwmode is set as 11a.
>>> I have also enable the following ht_capabilities on my radio in
>>> /etc/config/wireless:
>>>
>>>         list ht_capab 'short_gi_80'
>>>         list ht_capab 'max-mpdu-11454'
>>>         list ht_capab 'TX-STBC-2BY1'
>>>         list ht_capab 'RX-STBC1'
>>>         list ht_capab 'RX-ANTENNA-PATTERN'
>>>         list ht_capab 'TX-ANTENNA-PATTERN'
>>>         list ht_capab 'MAX-A-MPDU-LEN-EXP7'
>>>
>>> I just finished running iperf3 from my laptop (2nd gen Core i7) to 
>>> one of my nodes. The results were indeed a bit different.
>>> Iperf server was running on my node and my laptop was the client.
>>> UDP:
>>> Normal: average result of 221Mbit, which is pretty normal I think.
>>> Reverse: average result of 41Mbit, which is better than the average 
>>> 20Mbit I got when running iperf3 using 2 nodes.
>>> TCP:
>>> Normal: average result of 49Mbit
>>> Reverse: average result of 35.4Mbit
>>>
>>> In terms of congestion, I'm not sure what you mean, but I'm in a low 
>>> populated area with few wireless access points. I disabled all my 
>>> 5GHz devices except for the mesh nodes. InSSIDer only shows my nodes 
>>> in the 5GHz range.
>>>
>>> Is there anything else I could try for you?
>>>
>>> Met vriendelijke groeten,
>>> Ralph Houthoofdt
>>> Student industriële wetenschappen informatica
>>>
>>> -----Original Message-----
>>> From: ath10k [mailto:ath10k-bounces at lists.infradead.org] On Behalf 
>>> Of Ben Greear
>>> Sent: dinsdag 28 februari 2017 17:56
>>> To: Ralph Houthoofdt <ralph at rora.be>; ath10k at lists.infradead.org
>>> Subject: Re: QCA9880 slow performance adhoc/IBSS mode
>>>
>>> Run iperf on a separate machine, as the apu2 have too weak of a CPU 
>>> to drive the wifi network at full speed.  And, publish what results 
>>> you do get, including RF information such as channel and how 
>>> congested is your test area.  Try udp as well as tcp.
>>>
>>> Thanks,
>>> Ben
>>>
>>> On 02/28/2017 08:51 AM, Ralph Houthoofdt wrote:
>>>> Dear all,
>>>>
>>>> I’m currently doing me masters thesis on the subject of Wireless 
>>>> Mesh Networks and how to optimize them for home use.
>>>> The devices I use are 5x PC Engine APU2c4 with 2x QCA9880 chip 
>>>> (Compex WLE900VX). All of them are running LEDE (OpenWRT fork).
>>>>
>>>> The problem I’m experiencing is the same as 
>>>> http://lists.infradead.org/pipermail/ath10k/2017-January/009075.html .
>>>> Using the tips mentioned I did not come to a better performance. 
>>>> I’m thinking it’s driver related..
>>>> When running iperf3 only one direction is running at a reasonable 
>>>> throughput. Since I am using adhoc mode primarily I decided to use 
>>>> these
>>>> packages:
>>>> -	Kmod-ath10k-ct
>>>> -	Ath10k-firmware-qca988x-ct
>>>> Some sources on the internet said that using the Candela Tech 
>>>> drivers should result in better support for adhoc/mesh networks.
>>>>
>>>> As suggested in the other issue linked above, I switched TCP 
>>>> congestion control to ‘Reno’ instead of the default ‘Cubic’.
>>>> I also did try the command “iw reg set BE” since I am from Belgium, 
>>>> but running “iw reg get” right after still gave me US as a result.
>>>>
>>>> Let me know what additional command info I should post. Any help 
>>>> would be greatly appreciated.
>>>>
>>>> Iwconfig:
>>>> adhoc0    IEEE 802.11  ESSID:"mesh"
>>>>           Mode:Ad-Hoc  Frequency:5.18 GHz  Cell: 02:CA:FE:CA:CA:40
>>>>           Tx-Power=20 dBm
>>>>           RTS thr:off   Fragment thr:off
>>>>           Encryption key:off
>>>>           Power Management:off
>>>>
>>>>
>>>> Iw reg get (I’m from Belgium):
>>>> global
>>>> country BE: DFS-ETSI
>>>>         (2402 - 2482 @ 40), (N/A, 20), (N/A)
>>>>         (5170 - 5250 @ 80), (N/A, 20), (N/A), AUTO-BW
>>>>         (5250 - 5330 @ 80), (N/A, 20), (0 ms), DFS, AUTO-BW
>>>>         (5490 - 5710 @ 160), (N/A, 27), (0 ms), DFS
>>>>         (57000 - 66000 @ 2160), (N/A, 40), (N/A)
>>>>
>>>> phy#1
>>>> country US: DFS-FCC
>>>>         (2402 - 2472 @ 40), (N/A, 30), (N/A)
>>>>         (5170 - 5250 @ 80), (N/A, 23), (N/A), AUTO-BW
>>>>         (5250 - 5330 @ 80), (N/A, 23), (0 ms), DFS, AUTO-BW
>>>>         (5490 - 5730 @ 160), (N/A, 23), (0 ms), DFS
>>>>         (5735 - 5835 @ 80), (N/A, 30), (N/A)
>>>>         (57240 - 63720 @ 2160), (N/A, 40), (N/A)
>>>>
>>>> phy#0
>>>> country US: DFS-FCC
>>>>         (2402 - 2472 @ 40), (N/A, 30), (N/A)
>>>>         (5170 - 5250 @ 80), (N/A, 23), (N/A), AUTO-BW
>>>>         (5250 - 5330 @ 80), (N/A, 23), (0 ms), DFS, AUTO-BW
>>>>         (5490 - 5730 @ 160), (N/A, 23), (0 ms), DFS
>>>>         (5735 - 5835 @ 80), (N/A, 30), (N/A)
>>>>         (57240 - 63720 @ 2160), (N/A, 40), (N/A)
>>>>
>>>>
>>>> Kind regards,
>>>> Met vriendelijke groeten,
>>>> Ralph Houthoofdt
>>>> Student industriële wetenschappen informatica
>>>>
>>>>
>>>> _______________________________________________
>>>> ath10k mailing list
>>>> ath10k at lists.infradead.org
>>>> http://lists.infradead.org/mailman/listinfo/ath10k
>>>>
>>>
>>> --
>>> Ben Greear <greearb at candelatech.com> Candela Technologies Inc  
>>> http://www.candelatech.com
>>>
>>> _______________________________________________
>>> ath10k mailing list
>>> ath10k at lists.infradead.org
>>> http://lists.infradead.org/mailman/listinfo/ath10k
>>>
>>
>> --
>> Ben Greear <greearb at candelatech.com>
>> Candela Technologies Inc  http://www.candelatech.com
>>
>
> --
> Ben Greear <greearb at candelatech.com>
> Candela Technologies Inc  http://www.candelatech.com
>

--
Ben Greear <greearb at candelatech.com>
Candela Technologies Inc  http://www.candelatech.com




More information about the ath10k mailing list