RE: [ns] Strange DiffServ Results.

[Sungwon Yi <syi@cse.psu.edu>]

Hi, I am Sungwon.

If the type of the traffic that I used in simulation is TCP, I agree.
The packet drop in the core router can cause the queue for TCP traffic
empty for short period of time due to timeout.
In the sample script, 2 UDP of 4M rate are used for both classes.
therefore, I expected no change in the other class.

thanks for the reply again.

Sungwon.

======================================
Yi, Sungwon
Have a wonderful day!
======================================

On Mon, 16 Jul 2001, Sergio Andreozzi wrote:

> I think your results are correct. If you reduce one queue lenght, you reduce
> the queue latency, moreover you increase the number of packet dropped for
> buffer overflow.
>
> On the other hand you increase the probability the scheduler find this queue
> in idle status. In this case the other queue is served, more packets are
> sent and the dropping percentage decrease consequently.
>
> Do you agree?
>
>
> Sergio
>
>
> > -----Original Message-----
> > From: Sungwon Yi [mailto:syi@cse.psu.edu]
> > Sent: 16. heinakuuta 2001 21:01
> > To: Sergio Andreozzi
> > Cc: ns-users@ISI.EDU
> > Subject: Re: BingGo!!!
> >
> >
> > First of all, I really appreciate for your fast and kind response.
> >
> > Hi, I am sungwon, again.
> > What you suggested is exactly what i did.
> >
> > It seems that new queue size is assigned to the queue.
> > The thing is that I only changed one of two queues.
> > However, this modification also affected the number of
> > packets dropped in other queue though the size remains
> > the same. : tcl/ex/diffserv/ds-cbr-tb-WRR.tcl
> >
> > Could anyone explain why the change of one queue size
> > affect the number of drop in other queue( class ) ?
> >
> > Best Regards,
> > Sungwon.
> >
> > ======================================
> > Yi, Sungwon
> > Have a wonderful day!
> > ======================================
> >
> >
>
>