Re: Xgraph output Question

[Ahmed A-G Helmy <ahelmy@catarina.usc.edu>]

> Hi,
> 
>    I am running some ns-lan simulations. I believe in the default
>    which I am using, all queues are being traced. But it was hard for
>    me to interpret the graphs as they would come out very busy.
> 
>    Anyone knows of detailed documentations on these graphs?

Emad,

	I am no expert in xgraph myself (in fact I know so little about 
it, just to get by)... 
	but you can probably reduce the number of data points you feed 
into xgraph that would increase the graph's readability.. by maybe using 
some scripting filters or so.. 

	now I don't know the general setup of your simulation 
environment, but I usually filter out the trace files based on the class_ 
(called flow id field in C++) parameter..
for example, if I want to look at the packets sent by the PIM protocol 
(e.g. join, graft, assert,... etc) I would create an agent, such as:
	set prune [new Agent/Mcast/Prune/detailedDM $self]

	then when sending packets I set the class_ parameter based on the 
message type, for example:

	if { $which == "graft" } {
                $prune set class_ 31
        } else {
                # graft-ack
                $prune set class_ 32
        }        

[check tcl/mcast/detailedDM.tcl f u want]

now, this is probably a quick hack to do things... this field is not used 
in this way only... actually it should be used for other things.

probably one is able to extend the ns tracing capabilities to write out 
the packet types., protocol codes .. etc. but I haven't used such 
features before.

There are also other tricks by which you can trace the agents themselves 
instead of the links (I have implemented some of those for my work..),

hope this helps a bit,

Regards,
-A

> 
>   your response is appreciated?
> 
> Thanks,
> Emad.
> 
> 
>