[ns] RSVP-with-ns paper noted

[Lloyd Wood <l.wood@eim.surrey.ac.uk>]

http://www.cs.columbia.edu/~pingpan/paper_list.html

P. Pan and H. Schulzrinne, "Lightweight Resource Reservation
Signaling: Design, Performance and Implementation", Bell Labs
Technical Memorandum 10009669-03, July 2000.

'we used the ns simulator and its RSVP module, and extended it to
 support partial reservation functionality required for our
 experiments.'

L.

<L.Wood@surrey.ac.uk>PGP<http://www.ee.surrey.ac.uk/Personal/L.Wood/>

---------- Forwarded message ----------
Date: Mon, 7 Aug 2000 20:01:14 -0400 (EDT)
From: Ping Pan <pingpan@cs.columbia.edu>
To: Qingming Ma <qma@cisco.com>
Cc: Peter Ashwood-Smith <petera@nortelnetworks.com>, mpls <mpls@UU.NET>
Subject: Re: traces to simulate QoS requests


Hi,

To study RSVP behavior, you can also modify the ns-2 RSVP code. Please go
to http://www.cs.columbia.edu/~pingpan/paper_list.html for one of my 
recent papers on lightweight reservation mechanisms. I had simulated a
good sized network and generated on/off RSVP requests with different bw
requirements. The call duration is also adjustable. We had found some
interesting problems with RSVP blockade mechanism, and worked out some
enhancements.

BTW, I know we cannot model data flows in possion, but what's wrong with
possion arrival for reservation requests? 

Thanks!

- Ping Pan


On Mon, 7 Aug 2000, Qingming Ma wrote:

> In my dissertation work on QoS routing, I used various different traffic
> models for QoS session arrivals, bandwidth requirements, and call
> holding time distributions. You can check out details from my dissertation.
> Let me know if you want to have a copy.
> 
> Thank you,
> 
> Qingming
> 
> 
> At 08:56 AM 5/4/00 -0400, Peter Ashwood-Smith wrote:
> >MPLS wrote:
> > >
> > > Hi all,
> > > We are trying to simulate a workmodel for handling QoS requests.
> > >
> > > In this respect, are there any publicly available traces for real time
> > > call requests that request bandwidth at the ISPs.
> > >
> > > If not, is there some study that says what is a feasible
> > > distribution for the following:
> > > 1. b/w requirement of a QoS call
> > > 2. QoS call request arrival rate
> > > 3. QoS call duration
> > >
> > > Most of the reserach work has been done using uniform ditsribution
> > > for b/w requirement, poisson for arrival rate, and exponential for call
> > > duration.
> > > How close to the real life scenario are these assumptions.
> > >
> > > Thanks in advance,
> > >
> > > Gargi Banerjee
> >
> >   I'm not aware of a source of the data you are looking for however I'd
> >warn you against using any of the nice clean models like poisson etc.
> >You may be able to extrapolate from some existing ATM networks as a
> >starting point.
> >
> >   The rate that LSPs arrive, depart and how long they hold will depend
> >on the applications using them and we cannot easily predict what applications
> >will tend to do with them. Initially you will see REALLY long hold times
> >but these will come down in time as more and more uses for LSPs evolve.
> >
> >   My gut feeling is that LSPs will eventually have behavior like data only
> >at a slower rate. I.e. chaotic, bursty etc.
> >
> >   Cheers,
> >
> >   Peter