2016-02-07 04:21:46

by Nikolay Martynov

[permalink] [raw]
Subject: Monitor interface TX performance boost

Hi.

I'm seeing strange behavior on my intel 6300 card.

If I measure TX performance just after boot (iperf in tcp mode), I get this:

[ 3] 0.0-10.1 sec 11.2 MBytes 9.32 Mbits/sec

But if create and enable monitor interface (with command like 'sudo
iw phy phy0 interface add moni0 type monitor && sudo ifconfig moni0
up') I get this:

[ 3] 0.0-10.1 sec 34.8 MBytes 29.0 Mbits/sec

That's about three times boost in throughput by just creating an
interface (without even using it).
Doing 'ifconfig moni0 down' bring performance back to 10Mbits range.

Experiments are done with laptop being ~5m from router and not moving.
These results are repeatable, i.e. not caused by random things like
other stations transmitting, or channel being busy.

Is this something expected? Is there any reason performance can be as
good without monitor interface existing?

Thanks!

--
Martynov Nikolay.
Email: [email protected]


2016-02-07 07:27:40

by Emmanuel Grumbach

[permalink] [raw]
Subject: Re: Monitor interface TX performance boost

On Sun, Feb 7, 2016 at 9:09 AM, Kalle Valo <[email protected]> wrote:
> Nikolay Martynov <[email protected]> writes:
>
>> I'm seeing strange behavior on my intel 6300 card.
>>
>> If I measure TX performance just after boot (iperf in tcp mode), I get this:
>>
>> [ 3] 0.0-10.1 sec 11.2 MBytes 9.32 Mbits/sec
>>
>> But if create and enable monitor interface (with command like 'sudo
>> iw phy phy0 interface add moni0 type monitor && sudo ifconfig moni0
>> up') I get this:
>>
>> [ 3] 0.0-10.1 sec 34.8 MBytes 29.0 Mbits/sec
>>
>> That's about three times boost in throughput by just creating an
>> interface (without even using it).
>> Doing 'ifconfig moni0 down' bring performance back to 10Mbits range.
>>
>> Experiments are done with laptop being ~5m from router and not moving.
>> These results are repeatable, i.e. not caused by random things like
>> other stations transmitting, or channel being busy.
>>
>> Is this something expected? Is there any reason performance can be as
>> good without monitor interface existing?
>
> Maybe the monitor mode disables power save automatically?
>

That could be an option although I remember to have disabled power
save by default (cough..).
That happened in 3.17 IIRC.

2016-02-07 07:09:06

by Kalle Valo

[permalink] [raw]
Subject: Re: Monitor interface TX performance boost

Nikolay Martynov <[email protected]> writes:

> I'm seeing strange behavior on my intel 6300 card.
>
> If I measure TX performance just after boot (iperf in tcp mode), I get this:
>
> [ 3] 0.0-10.1 sec 11.2 MBytes 9.32 Mbits/sec
>
> But if create and enable monitor interface (with command like 'sudo
> iw phy phy0 interface add moni0 type monitor && sudo ifconfig moni0
> up') I get this:
>
> [ 3] 0.0-10.1 sec 34.8 MBytes 29.0 Mbits/sec
>
> That's about three times boost in throughput by just creating an
> interface (without even using it).
> Doing 'ifconfig moni0 down' bring performance back to 10Mbits range.
>
> Experiments are done with laptop being ~5m from router and not moving.
> These results are repeatable, i.e. not caused by random things like
> other stations transmitting, or channel being busy.
>
> Is this something expected? Is there any reason performance can be as
> good without monitor interface existing?

Maybe the monitor mode disables power save automatically?

--
Kalle Valo