* Re: [ath10k throughput] low throughput in multi-user mode
@ 2018-03-21 16:31 ` Ben Greear
0 siblings, 0 replies; 6+ messages in thread
From: Ben Greear @ 2018-03-21 16:31 UTC (permalink / raw)
To: gary, linux-wireless, ath10k
On 03/20/2018 06:44 PM, gary wrote:
>
> Hi all,
> I have run the throughput test on veriwave.(from Ethernet to wireless)
> My AP suports 4*4, 11ac, mu-mimo.,wireless chip is QCA9984.
> At first, the throughput is about 80Mbps only with 16 users, so I force to
> return true in function ath10k_mac_tx_can_push().
> As a result, when the user number is 1~16, the udp throughput is about
> 1.2Gbps ~1.4Gbps.
> But when the user number is 32, the udp throughput is only 520Mbps.
>
> So I try to substitute ath10k-firmware.
> With firmware 10.4-3.4-0082 and above, the throughput is around 520Mbps.
> But with firmware 10.4-3.4-0072, the throughput gets 1.1Gbps.
>
> Do I miss something in configuration?
> Any comment is appreciated.
What kernel version (or backports, if you are using that?) are you using?
And, I'd be curious how the latest ath10k-CT (beta) firmware compares
if you have time to try that:
http://www.candelatech.com/ath10k-10.4.php
At least in upload testing, we get better throughput with lots of
virtual stations if we tune the number of tx descriptors to be larger
as that will generate larger AMPDUs on air.
You might also check your rate-ctrl logic to make sure all of your
stations are using higher MCS rates.
Thanks,
Ben
--
Ben Greear <greearb@candelatech.com>
Candela Technologies Inc http://www.candelatech.com
^ permalink raw reply [flat|nested] 6+ messages in thread
* RE: [ath10k throughput] low throughput in multi-user mode
2018-03-21 16:31 ` Ben Greear
(?)
@ 2018-03-23 3:04 ` gary
-1 siblings, 0 replies; 6+ messages in thread
From: gary @ 2018-03-23 3:04 UTC (permalink / raw)
To: 'Ben Greear', linux-wireless, ath10k
Hi Ben,
Thank you for your reply.
My wireless driver is compat-wireless_2017-10-6, and kernel is 4.9.58.
The firmware can't work on my testbed.
By my observation, in one try, ath10k_htt_rx_tx_mode_switch_ind is invoked
only twice.
As a result, the ath10k driver can't get the latest num_push_allowed(always
zero).
Do you have any idea of it?
-----Original Message-----
From: ath10k [mailto:ath10k-bounces@lists.infradead.org] On Behalf Of Ben
Greear
Sent: 2018年3月22日 0:31
To: gary; linux-wireless@vger.kernel.org; ath10k@lists.infradead.org
Subject: Re: [ath10k throughput] low throughput in multi-user mode
On 03/20/2018 06:44 PM, gary wrote:
>
> Hi all,
> I have run the throughput test on veriwave.(from Ethernet to wireless)
> My AP suports 4*4, 11ac, mu-mimo.,wireless chip is QCA9984.
> At first, the throughput is about 80Mbps only with 16 users, so I
> force to return true in function ath10k_mac_tx_can_push().
> As a result, when the user number is 1~16, the udp throughput is about
> 1.2Gbps ~1.4Gbps.
> But when the user number is 32, the udp throughput is only 520Mbps.
>
> So I try to substitute ath10k-firmware.
> With firmware 10.4-3.4-0082 and above, the throughput is around 520Mbps.
> But with firmware 10.4-3.4-0072, the throughput gets 1.1Gbps.
>
> Do I miss something in configuration?
> Any comment is appreciated.
What kernel version (or backports, if you are using that?) are you using?
And, I'd be curious how the latest ath10k-CT (beta) firmware compares if you
have time to try that:
http://www.candelatech.com/ath10k-10.4.php
At least in upload testing, we get better throughput with lots of virtual
stations if we tune the number of tx descriptors to be larger as that will
generate larger AMPDUs on air.
You might also check your rate-ctrl logic to make sure all of your stations
are using higher MCS rates.
Thanks,
Ben
--
Ben Greear <greearb@candelatech.com>
Candela Technologies Inc http://www.candelatech.com
_______________________________________________
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
^ permalink raw reply [flat|nested] 6+ messages in thread
* RE: [ath10k throughput] low throughput in multi-user mode
2018-03-21 16:31 ` Ben Greear
(?)
(?)
@ 2018-03-23 3:21 ` gary
2018-03-23 16:06 ` Ben Greear
-1 siblings, 1 reply; 6+ messages in thread
From: gary @ 2018-03-23 3:21 UTC (permalink / raw)
To: 'Ben Greear', ath10k
Hi Ben,
Thank you for your reply.
My wireless driver is compat-wireless_2017-10-6, and kernel is 4.9.58.
The test result with your firmware:
Get throughput 1.13Gbps with 16 users.
And get almost 0 with 32 users.
By my observation, in one try, ath10k_htt_rx_tx_mode_switch_ind is invoked
only twice.
As a result, the ath10k driver can't get the latest num_push_allowed(always
zero).
Do you have any idea of it?
-----Original Message-----
From: ath10k [mailto:ath10k-bounces@lists.infradead.org] On Behalf Of Ben
Greear
Sent: 2018年3月22日 0:31
To: gary; linux-wireless@vger.kernel.org; ath10k@lists.infradead.org
Subject: Re: [ath10k throughput] low throughput in multi-user mode
On 03/20/2018 06:44 PM, gary wrote:
>
> Hi all,
> I have run the throughput test on veriwave.(from Ethernet to wireless)
> My AP suports 4*4, 11ac, mu-mimo.,wireless chip is QCA9984.
> At first, the throughput is about 80Mbps only with 16 users, so I
> force to return true in function ath10k_mac_tx_can_push().
> As a result, when the user number is 1~16, the udp throughput is about
> 1.2Gbps ~1.4Gbps.
> But when the user number is 32, the udp throughput is only 520Mbps.
>
> So I try to substitute ath10k-firmware.
> With firmware 10.4-3.4-0082 and above, the throughput is around 520Mbps.
> But with firmware 10.4-3.4-0072, the throughput gets 1.1Gbps.
>
> Do I miss something in configuration?
> Any comment is appreciated.
What kernel version (or backports, if you are using that?) are you using?
And, I'd be curious how the latest ath10k-CT (beta) firmware compares if you
have time to try that:
http://www.candelatech.com/ath10k-10.4.php
At least in upload testing, we get better throughput with lots of virtual
stations if we tune the number of tx descriptors to be larger as that will
generate larger AMPDUs on air.
You might also check your rate-ctrl logic to make sure all of your stations
are using higher MCS rates.
Thanks,
Ben
--
Ben Greear <greearb@candelatech.com>
Candela Technologies Inc http://www.candelatech.com
_______________________________________________
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
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [ath10k throughput] low throughput in multi-user mode
2018-03-23 3:21 ` gary
@ 2018-03-23 16:06 ` Ben Greear
0 siblings, 0 replies; 6+ messages in thread
From: Ben Greear @ 2018-03-23 16:06 UTC (permalink / raw)
To: gary, ath10k
On 03/22/2018 08:21 PM, gary wrote:
>
> Hi Ben,
>
> Thank you for your reply.
> My wireless driver is compat-wireless_2017-10-6, and kernel is 4.9.58.
>
> The test result with your firmware:
> Get throughput 1.13Gbps with 16 users.
> And get almost 0 with 32 users.
Can you try an older kernel w/out backports, maybe my 4.13 or 4.9, or
some stock kernel that is a bit older?
https://github.com/greearb/linux-ct-4.9
Thanks,
Ben
>
> By my observation, in one try, ath10k_htt_rx_tx_mode_switch_ind is invoked
> only twice.
> As a result, the ath10k driver can't get the latest num_push_allowed(always
> zero).
>
> Do you have any idea of it?
>
> -----Original Message-----
> From: ath10k [mailto:ath10k-bounces@lists.infradead.org] On Behalf Of Ben
> Greear
> Sent: 2018年3月22日 0:31
> To: gary; linux-wireless@vger.kernel.org; ath10k@lists.infradead.org
> Subject: Re: [ath10k throughput] low throughput in multi-user mode
>
>
> On 03/20/2018 06:44 PM, gary wrote:
>>
>> Hi all,
>> I have run the throughput test on veriwave.(from Ethernet to wireless)
>> My AP suports 4*4, 11ac, mu-mimo.,wireless chip is QCA9984.
>> At first, the throughput is about 80Mbps only with 16 users, so I
>> force to return true in function ath10k_mac_tx_can_push().
>> As a result, when the user number is 1~16, the udp throughput is about
>> 1.2Gbps ~1.4Gbps.
>> But when the user number is 32, the udp throughput is only 520Mbps.
>>
>> So I try to substitute ath10k-firmware.
>> With firmware 10.4-3.4-0082 and above, the throughput is around 520Mbps.
>> But with firmware 10.4-3.4-0072, the throughput gets 1.1Gbps.
>>
>> Do I miss something in configuration?
>> Any comment is appreciated.
>
> What kernel version (or backports, if you are using that?) are you using?
>
> And, I'd be curious how the latest ath10k-CT (beta) firmware compares if you
> have time to try that:
>
> http://www.candelatech.com/ath10k-10.4.php
>
> At least in upload testing, we get better throughput with lots of virtual
> stations if we tune the number of tx descriptors to be larger as that will
> generate larger AMPDUs on air.
>
> You might also check your rate-ctrl logic to make sure all of your stations
> are using higher MCS rates.
>
> Thanks,
> Ben
>
> --
> Ben Greear <greearb@candelatech.com>
> Candela Technologies Inc http://www.candelatech.com
>
>
> _______________________________________________
> 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
>
--
Ben Greear <greearb@candelatech.com>
Candela Technologies Inc http://www.candelatech.com
_______________________________________________
ath10k mailing list
ath10k@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/ath10k
^ permalink raw reply [flat|nested] 6+ messages in thread