[c-nsp] ipsla - latency - related to cellular backhaul

[Pete Lumbis]

You are correct. The ICMP Echo Reply would be generated by the LC CPU on a
GSR. The CPU is slower, and yes there could still be non-deterministic
jitter, you sill are not going through the OS scheduler on the RP CPU like
you would on any other box.

For the ASR1k, the CPU is on the RP and the forwarding process is the ESP.
The Echo Reply is handled by one of the cores on the ESP. These cores are
also responsible for normal packet processing, so you'll see significantly
less jitter in this case.


On Sat, Apr 27, 2013 at 8:13 AM, Saku Ytti <saku at ytti.fi> wrote:

> On (2013-04-26 09:55 -0400), Pete Lumbis wrote:
>
> > Some hardware platforms and offload ping, mainly Echo Reply. I know that
> > ASR1k and the GSR can do this off the top of my head (that is, I'm not
> > saying this is an exhaustive list). Echo Requests will always be
> generated
> > by the Supervisor/RP/Central CPU.
>
> I'm not very familiar with ASR1k. But I believe if GSR does something
> special to them, it handles them in LC CPU, which is even lower performance
> CPU than RP CPU. I don't believe it has capability to handle them in LC
> forwarding HW.
> Of course it's still better than handling in LC CPU + RP CPU, but it's
> still significant cause of jitter.
>
> Also, I'm going to plug co-workers IP SLA/JNPR RPM responder for linux,
> it's quite easy/cheap way to reduce jitter cause by responder.
> Here's result happy user got from switching ASR1k responder to it:
>
> http://cmouse.github.io/ip-sla-responder/testimonial_01.html
>
> --
>   ++ytti
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/
>