Is there a way to instruct NIC/DPDK which RXQ to deliver a packet to not RSS?

Is there a way to instruct NIC/DPDK which RXQ to deliver a packet to not RSS?

[fwefew 4t4tg]

[-- Attachment #1: Type: text/plain, Size: 294 bytes --]

I am aware of RSS; I coded it and have gotten it to work between machines.

However, the elapsed time per packet increases from <100ns/packet RSS-OFF
to 700-800ns RSS-ON in my test setup.

In my scenario I have hardcoded routing between TXQs on client machines and
RXQs on the server machines.

[-- Attachment #2: Type: text/html, Size: 371 bytes --]

Re: Is there a way to instruct NIC/DPDK which RXQ to deliver a packet to not RSS?

[Stephen Hemminger]

On Sun, 18 Dec 2022 16:41:13 -0500
fwefew 4t4tg <7532yahoo@gmail.com> wrote:

> I am aware of RSS; I coded it and have gotten it to work between machines.
> 
> However, the elapsed time per packet increases from <100ns/packet RSS-OFF
> to 700-800ns RSS-ON in my test setup.
> 
> In my scenario I have hardcoded routing between TXQs on client machines and
> RXQs on the server machines.

Packet direction to queue in HW is done with rte_flow (if your NIC supports it)
https://doc.dpdk.org/guides/prog_guide/rte_flow.html

RSS should be implemented in hardware. So it shouldn't be causing added
latency. What device are you using.

Re: Is there a way to instruct NIC/DPDK which RXQ to deliver a packet to not RSS?

[fwefew 4t4tg]

[-- Attachment #1: Type: text/plain, Size: 1133 bytes --]

>RSS should be implemented in hardware. So it shouldn't be causing added
>latency. What device are you using.
My error; aggregate performance is the same. H/W RSS is fine.

>Packet direction to queue in HW is done with rte_flow (if your NIC
supports it)
>https://doc.dpdk.org/guides/prog_guide/rte_flow.html
Thank you! That was not on my radar for some reason. Much appreciated.


On Sun, Dec 18, 2022 at 5:14 PM Stephen Hemminger <
stephen@networkplumber.org> wrote:

> On Sun, 18 Dec 2022 16:41:13 -0500
> fwefew 4t4tg <7532yahoo@gmail.com> wrote:
>
> > I am aware of RSS; I coded it and have gotten it to work between
> machines.
> >
> > However, the elapsed time per packet increases from <100ns/packet RSS-OFF
> > to 700-800ns RSS-ON in my test setup.
> >
> > In my scenario I have hardcoded routing between TXQs on client machines
> and
> > RXQs on the server machines.
>
> Packet direction to queue in HW is done with rte_flow (if your NIC
> supports it)
> https://doc.dpdk.org/guides/prog_guide/rte_flow.html
>
> RSS should be implemented in hardware. So it shouldn't be causing added
> latency. What device are you using.
>

[-- Attachment #2: Type: text/html, Size: 1810 bytes --]