[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Node throughput and packet delay



[Hope I'm not bugging ns-users with my rantings - I think it's useful to
everyone.]

Tarik,

> 	i agree with you: dumping everything to the tracefile is an easy 
> solution and it takes a lot of space. 
> 	However there are flow monitor objects that can be used to collect 
> statistics without much postprocessing.

Sure - these are something that I didn't really pick up on, when I first
ventured into the world of ns. They are certainly useful, but I think
they more of such functionality can be added.

> 	I also agree on the fact that if you want to collect specific data, well 
> you may as well derive your own monitor object. But still if you are a new user 
> it might just be easier to use trace-all and postprocess with awk than go  on 
> the darkside of ns: C++. I think that working on the C++ part of ns requires a 
> strong knowledge of ns internals.

This is not really what I was saying - my opinion is that it would
probably be useful to have some more monitoring objects provided as
standard in ns, that generate common results. People could use these at
the otcl level - they need not look under the hood and get involved in c++
stuff. Of course, if you have some very specific purpose, for which you
need highly unusual results then of course you can't expect ns to provide
them. But in the main, I think that many people want to obtain similar
types of results, some of which ns can already output very easily, some
of which are easily obtained, but ns does not have functionality to do
this as standard.

So, for example, there have been a few posts to the list by new users
wanting to know how to obtain goodput info - Haobo Yu posted a Tcl class
that can be used to obtain the goodput obtained by a particular source. I
think it would be good if a c++ variant of this was included in ns (I
think the tcl version is a bit slow). Ideally, this should also be
referenced in the tutorial.

Packet delays and variations are also something that I'm sure people need
to obtain reasonably frequently, and, again, I think it would be good if
there was functionality in ns to do this, rather than having every user
devise a solution (be it in tcl/c++ or perl or awk) themselves.

It seems that there are solutions to these problems in the software
supplied with ns in the form of awk and perl scripts, but it's something I
never looked into - perhaps I missed something in the documentation.
However, I still think that it would be better if such functionality was
more integrated into ns itself, rather than in ancilliary scripts.

This would be good you say (perhaps) - there are lots of things that would
be good - am I willing to do anything about it? Well, the two types of
monitoring component that I referred to above are really quite simple to
implement - I'd be willing to do this - I've already done it, but if it
was to be put into ns, then I'd need to polish up my stuff a wee bit. The
key thing would be to produce some documentation/examples, which again I
could do.

Sean.
-----
Sean Murphy,			Email: [email protected]
Teltec Ireland,			Phone: +353-1-7045080
DCU, Dublin 9,			Fax:   +353-1-7045092
Ireland.