Re: LLQ- help (with elephants)

Let's go back to being technical, shall we?

On Wed, Dec 19, 2012 at 4:28 PM, Paul Negron <negron.paul_at_gmail.com> wrote:
>
> I said it DID NOT congest the link. the total bandwidth was under 128 K at
> that point. Voice is very SOLID. Not bursty at all.

If 4 calls make up 112 Kb/s, then 5 calls make up 140 Kb/s, given the
same codec. That should have exceeded 128 Kb/s. There are two
explanations:

1. Calls were timed such that packets interleaved and never triggered
a congestion
2. Burst parameter actually does what Cisco documentation says it does
and it allows the traffic to exceed the rate temporarily (while the
link is congested). At this point, AGAIN, I am at loss:

- You and Narbik seem to claim *different* things about burst. He
claims (in the previous message and the lab he posted) that it works
like Bc. Fair enough.
- However, your own notes and findings seem to indicate that it
actually works similar to what be does with a two rate, three color
policer.

Really... at this point... Which one is it?

> The burst can be configured to help drop more aggressively dude!!! Or extend
> to not drop packets if need be.

Well... in which cases? When the traffic is conforming to the rate,
exceeding? We all agree this happens when LLQ is engaged, so let's
focus on when the burst does... the above (again somewhat
contradictory to both your previous statements and Narbik's).

> WHY do we configure Q limit then? So the drop happens somewhere else?

The traffic is still in the class, it's just not being transmitted.
It's never reclassified. There's no recursive classification, that I'm
aware of.

> You are taking me out of context here. I said the burst was high for the
> rate being sent through the class but the burst parameter was not set to
> sustain it. That is what I inferred.

I don't understand what you're saying here. Which burst are you
referring to?! The parameter or something else - the temporary surge
in traffic?

> Last but not the least...Paul Negron

 ?

--
Marko Milivojevic - CCIE #18427 (SP R&S)
Senior CCIE Instructor - IPexpert
Blogs and organic groups at http://www.ccie.net