All of lore.kernel.org
 help / color / mirror / Atom feed
From: "Iñaki Pascual" <ipascual@cttc.cat>
To: ath10k@lists.infradead.org
Subject: Re: low TCP throughput with 80MHz channel width [SOLVED]
Date: Mon, 25 Apr 2016 09:48:16 +0200	[thread overview]
Message-ID: <571DCBC0.1010504@cttc.cat> (raw)
In-Reply-To: <571A57DC.9020101@candelatech.com>

Thanks Ben, Michal for your help.

It turns out that the "reno" thing fixed the issue.

We still do not know why it didn't work the first time but after some 
debbuging, but no config changes, it just started working.

We are now on 200 Mbps TCP on 80 MHz which makes sense.

Thanks again. Bests,

Iñaki

On 22/04/16 18:57, Ben Greear wrote:
> On 04/22/2016 09:44 AM, Iñaki Pascual wrote:
>> Thanks for your help Ben.
>>
>> I think we will go for kernel 4.4 and the latest firmware relase.
>>
>> Regarding the firmware, should we try the CT-10.2 or it is not stable 
>> enough?
>
> My 10.2 has not had much testing, and in recent testing, it does not 
> perform as well as my 10.1
> firmware.  But, you could give it a try to see how well it works....
>
> 4.4 seems to be working this morning, but was broken yesterday: It is 
> in fairly heavy development, so possibly you will
> hit regressions there.  Let me know if you do.  It also has a bunch of 
> upstream testing ath10k patches, and I saw other reports
> that TCP throughput is less with them applied, though possibly this is 
> only true if you are using codel tx queueing.
>
> Thanks,
> Ben
>
>>
>> Bests,
>>
>> Iñaki
>>
>>
>>
>> On 22/04/16 18:16, Ben Greear wrote:
>>> On 04/22/2016 09:11 AM, Iñaki Pascual wrote:
>>>> We are using kernel 4.2 with CandelaTech firmware 
>>>> 10.1.467-ct-com-full-014-96d543
>>>
>>> Maybe try rls 15 or beta-16 CT 10.1 firmware?
>>>
>>> And, we found too much instability in 4.2 kernel.  We suggest 4.0 or 
>>> 4.4,
>>> though I am not aware of any throughput related bugs aside from the 
>>> CUBIC
>>> congestion control issue.
>>>
>>> You might also try our 3.17 kernel since CUBIC worked in that kernel...
>>>
>>> Thanks,
>>> Ben
>>>
>>>
>>>>
>>>> monet@sc227:~$ ethtool -i wlan2
>>>> driver: ath10k_pci
>>>> version: 4.2.0+
>>>> firmware-version: 10.1.467-ct-com-full-014-96d543
>>>> bus-info: 0000:0b:00.0
>>>> supports-statistics: yes
>>>> supports-test: no
>>>> supports-eeprom-access: no
>>>> supports-register-dump: no
>>>> supports-priv-flags: no
>>>>
>>>> Thanks,
>>>>
>>>> Iñaki
>>>>
>>>> On 22/04/16 18:06, Ben Greear wrote:
>>>>> On 04/22/2016 09:02 AM, Iñaki Pascual wrote:
>>>>>
>>>>>> Thanks for the tip Ben. We have moved to 'reno' and tried several 
>>>>>> window sizes but we are still on max. 35 Mbps which is too low.
>>>>>>
>>>>>
>>>>> What firmware and kernel are you using?
>>>>>
>>>>> Thanks,
>>>>> Ben
>>>>>
>>>>>> Bests,
>>>>>>
>>>>>> Iñaki
>>>>>>
>>>>>> On 22/04/16 17:40, Ben Greear wrote:
>>>>>>> Use 'reno' instead of cubic TCP congestion control, cubic is 
>>>>>>> broken with ath10k.
>>>>>>>
>>>>>>> Thanks,
>>>>>>> Ben
>>>>>>>
>>>>>>> On 04/22/2016 08:32 AM, Iñaki Pascual wrote:
>>>>>>>> Hi everyone,
>>>>>>>>
>>>>>>>> we have configured two boxes with 802.11ac NICs to work in 
>>>>>>>> channel 48 (5240 MHz), ad-hoc mode, 80MHz channel width.
>>>>>>>> For UDP traffic we get 300 Mbps bandwith but for TCP we are 
>>>>>>>> around 30 Mbps.
>>>>>>>>
>>>>>>>> We have tried other channels with similar results.
>>>>>>>> Throughput for 20 and 40 MHz channel width are ok (20MHz: TCP 
>>>>>>>> 21Mbps, UDP 27Mbps, 40MHz: TCP 82Mbps, UDP 144Mbps).
>>>>>>>>
>>>>>>>> Testing with iperf we have also tried different congestion 
>>>>>>>> windows with no or low improvements.
>>>>>>>>
>>>>>>>> Below are config details and the sequence of commands we are 
>>>>>>>> using to configure the interface.
>>>>>>>>
>>>>>>>> Any ideas on what we may be doing wrong?
>>>>>>>>
>>>>>>>> Bests,
>>>>>>>>
>>>>>>>> Iñaki
>>>>>>>>
>>>>>>>> iw outputs:
>>>>>>>>
>>>>>>>> root@sc227:/home/monet# iw wlan2 info
>>>>>>>> Interface wlan2
>>>>>>>>      ifindex 6
>>>>>>>>      wdev 0x200000001
>>>>>>>>      addr 04:f0:21:18:26:a2
>>>>>>>>      ssid SC226SC227
>>>>>>>>      type IBSS
>>>>>>>>      wiphy 2
>>>>>>>>      channel 48 (5240 MHz), width: 80 MHz, center1: 5210 MHz
>>>>>>>>
>>>>>>>> root@sc227:/home/monet# iw wlan2 station dump
>>>>>>>> Station 04:f0:21:18:26:a8 (on wlan2)
>>>>>>>>      inactive time:    86 ms
>>>>>>>>      rx bytes:    467732521
>>>>>>>>      rx packets:    364249
>>>>>>>>      tx bytes:    1394558
>>>>>>>>      tx packets:    16200
>>>>>>>>      tx retries:    0
>>>>>>>>      tx failed:    0
>>>>>>>>      signal:      -51 dBm
>>>>>>>>      signal avg:    -51 dBm
>>>>>>>>      tx bitrate:    351.0 MBit/s VHT-MCS 4 80MHz VHT-NSS 2
>>>>>>>>      rx bitrate:    526.5 MBit/s VHT-MCS 4 80MHz VHT-NSS 3
>>>>>>>>      authorized:    yes
>>>>>>>>      authenticated:    yes
>>>>>>>>      preamble:    long
>>>>>>>>      WMM/WME:    yes
>>>>>>>>      MFP:        no
>>>>>>>>      TDLS peer:    no
>>>>>>>>      connected time:    3169 seconds
>>>>>>>>
>>>>>>>>
>>>>>>>> config commands:
>>>>>>>>
>>>>>>>> ifconfig wlan2 down
>>>>>>>> iw wlan2 set type ibss
>>>>>>>> ifconfig wlan2 up
>>>>>>>> ifconfig wlan2 10.2.1.7 netmask 255.255.255.0
>>>>>>>> iw wlan2 ibss join SC226SC227 5240 80MHz
>>>>>>>>
>>>>>>>> iperf:
>>>>>>>> root@sc227:/home/monet# iperf -c 10.2.1.6
>>>>>>>> ------------------------------------------------------------
>>>>>>>> Client connecting to 10.2.1.6, TCP port 5001
>>>>>>>> TCP window size: 85.0 KByte (default)
>>>>>>>> ------------------------------------------------------------
>>>>>>>> [  3] local 10.2.1.7 port 48548 connected with 10.2.1.6 port 5001
>>>>>>>> [ ID] Interval       Transfer     Bandwidth
>>>>>>>> [  3]  0.0-10.0 sec  33.1 MBytes  27.7 Mbits/sec
>>>>>>>>
>>>>>>>> _______________________________________________
>>>>>>>> ath10k mailing list
>>>>>>>> ath10k@lists.infradead.org
>>>>>>>> http://lists.infradead.org/mailman/listinfo/ath10k
>>>>>>>
>>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>
>>>
>>>
>>
>
>


_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k

  reply	other threads:[~2016-04-25  7:50 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-04-22 15:32 low TCP throughput with 80MHz channel width Iñaki Pascual
2016-04-22 15:40 ` Ben Greear
2016-04-22 16:02   ` Iñaki Pascual
2016-04-22 16:06     ` Ben Greear
2016-04-22 16:11       ` Iñaki Pascual
2016-04-22 16:16         ` Ben Greear
2016-04-22 16:44           ` Iñaki Pascual
2016-04-22 16:57             ` Ben Greear
2016-04-25  7:48               ` Iñaki Pascual [this message]
2016-04-25  7:07     ` Michal Kazior

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=571DCBC0.1010504@cttc.cat \
    --to=ipascual@cttc.cat \
    --cc=ath10k@lists.infradead.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.