[j-nsp] PLP Bit

Saku Ytti saku at ytti.fi
Sat Feb 20 08:24:51 EST 2016


On 20 February 2016 at 14:48,  <marco at lamehost.it> wrote:

Hey,

> This is a very basic question, but i can't find a clear statement about.
>
> As per my understanding JunOS has two different congestion avoidance
> mechanisms: RED and WRED.
> And the main difference is that RED relies only on PLP, while WRED let you
> drop at different rates on a per "color" basis.
>
> Is it right?
>
> In that case, how does PLP influence WRED operations?

PLP or packet loss priority is way to give multiple (W)RED curves for
single traffic class.

So you could specify one DSCP value to be AF low loss, and another
DSCP value to be AF high loss, you'd schedule them together in single
AF, give same buffer and same guaranteed percentage of traffic rate,
but you could give them different W()RED curves.

You could for example define two points for AF low loss curve,  0%
loss at 70% buffer utilisation and 100% loss at 100% buffer
utilisation. And for high loss you could say 0% loss at 20%
utilisation, 100% loss at 70% utilisation.

W in WRED means just that there is some discrimination on curves, that
not all traffic is subject to same drop probability at given buffer
utilisation.

-- 
  ++ytti


More information about the juniper-nsp mailing list