Re: [ns] Spaceway NGSO's simulation

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

On Wed, 16 May 2001, Tarik Taleb wrote:

> Hi everybody,
> I would like to simulate TCP and UDP traffic on certain satellite
> constellations. I have gotten the 

'a'. you can quibble over the designs till the cows come home
(and we did).

> simulation of Iridium and Teledesic
> from TCL/ex, and i need that of Spaceway NGSO.

I wrote that one myself; it's not (yet) part of the distribution.
Hacking the Teledesic script to create something that does a rosette
and is NGSO-equivalent is straightforward; you can find the orbital
parameters described in the SaVi script at:
http://www.ee.surrey.ac.uk/Personal/L.Wood/software/SaVi/

However, to support diversity management and surface control (the AIAA
paper below) the NGSO script needs to rely on a number of changes to
the handoff routine, themselves not yet checked into ns (because they
bring a number of things together rather inelegantly; lots of crufty
casting between classes that makes me think handover needs to be
reworked.) Without those modifications, you're at the mercy of
uncontrolled rosette handoff, and path delay fluctuates widely.


references talking about Spaceway NGSO simulation:

Managing diversity with handover to provide classes of service in
satellite constellation networks 
       Lloyd Wood, George Pavlou, Barry Evans
       conference paper, in the proceedings of the AIAA International
       Communication Satellite Systems Conference (ICSSC '01), vol. 3,
       session 35, no. 194, Toulouse, France, 17-20 April 2001.


Effects on TCP of routing strategies in satellite constellations 
       Lloyd Wood, George Pavlou, Barry Evans
       refereed journal paper, Special issue on Satellite-Based
       Internet Technology and Services, IEEE Communications Magazine,
       vol. 39, no. 3, pp. 172-181, March 2001. 


To simulate TCP traffic across a constellation I recommend building a
static model of the subset of the constellation you wish to run
traffic over, using static links, and neglecting orbital motion for
the (short) duration of the TCP flow. If you don't do that simulation
time and size will become a serious obstacle to you, and you'll start
worrying about the realism of transient effects during handover.
I ended up generating static topologies, then stripping them down...

L.

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