[LibreQoS] DPDK vs XDP

[LibreQoS] DPDK vs XDP

[Dave Taht]

I still get the giggles when I see the CPU usage on fig 4.

https://github.com/xdp-project/xdp-paper/blob/master/xdp-the-express-data-path.pdf

DPDK is a great way to heat data centers.

-- 
This song goes out to all the folk that thought Stadia would work:
https://www.linkedin.com/posts/dtaht_the-mushroom-song-activity-6981366665607352320-FXtz
Dave Täht CEO, TekLibre, LLC

Re: [LibreQoS] DPDK vs XDP

[Robert Chacón]

[-- Attachment #1: Type: text/plain, Size: 952 bytes --]

If we can get the IP filtering limit upped from 32k and introduce virtual
interfaces to split up HTBs as Toke suggested, we could see some crazy
throughput thanks to XDP. Very exciting!

On Thu, Oct 13, 2022 at 4:41 PM Dave Taht via LibreQoS <
libreqos@lists.bufferbloat.net> wrote:

> I still get the giggles when I see the CPU usage on fig 4.
>
>
> https://github.com/xdp-project/xdp-paper/blob/master/xdp-the-express-data-path.pdf
>
> DPDK is a great way to heat data centers.
>
> --
> This song goes out to all the folk that thought Stadia would work:
>
> https://www.linkedin.com/posts/dtaht_the-mushroom-song-activity-6981366665607352320-FXtz
> Dave Täht CEO, TekLibre, LLC
> _______________________________________________
> LibreQoS mailing list
> LibreQoS@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/libreqos
>


-- 
Robert Chacón
CEO | JackRabbit Wireless LLC <http://jackrabbitwireless.com>

[-- Attachment #2: Type: text/html, Size: 1806 bytes --]

Re: [LibreQoS] DPDK vs XDP

[Herbert Wolverson]

[-- Attachment #1: Type: text/plain, Size: 1693 bytes --]

Sorry I had to run - daycare schedule is tight if I want to park anywhere
remotely nearby.

I glanced through, and raising the hashmap/trie sizes is pretty
straightforward. The bigger problem looks to be the class numbering scheme
(major:minor). I wonder if it's possible to change the representation to
use 8 bits for Major/CPU and 24 for the minor? That would give us 16
million identifiers.

I'll try to research tomorrow.

On Thu, Oct 13, 2022, 5:48 PM Robert Chacón via LibreQoS <
libreqos@lists.bufferbloat.net> wrote:

> If we can get the IP filtering limit upped from 32k and introduce virtual
> interfaces to split up HTBs as Toke suggested, we could see some crazy
> throughput thanks to XDP. Very exciting!
>
> On Thu, Oct 13, 2022 at 4:41 PM Dave Taht via LibreQoS <
> libreqos@lists.bufferbloat.net> wrote:
>
>> I still get the giggles when I see the CPU usage on fig 4.
>>
>>
>> https://github.com/xdp-project/xdp-paper/blob/master/xdp-the-express-data-path.pdf
>>
>> DPDK is a great way to heat data centers.
>>
>> --
>> This song goes out to all the folk that thought Stadia would work:
>>
>> https://www.linkedin.com/posts/dtaht_the-mushroom-song-activity-6981366665607352320-FXtz
>> Dave Täht CEO, TekLibre, LLC
>> _______________________________________________
>> LibreQoS mailing list
>> LibreQoS@lists.bufferbloat.net
>> https://lists.bufferbloat.net/listinfo/libreqos
>>
>
>
> --
> Robert Chacón
> CEO | JackRabbit Wireless LLC <http://jackrabbitwireless.com>
> _______________________________________________
> LibreQoS mailing list
> LibreQoS@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/libreqos
>

[-- Attachment #2: Type: text/html, Size: 3140 bytes --]