[Ns-developers] flow-monitor features: time variant scenarios

[Sebastian Rohde]

Did you consider (optionally) keeping statistically data for multiple “time-slots” of configurable length in the flow monitor itself instead of putting the logic into the application via a reset of the flow monitor and periodically processing intermediate results? Or do you assume that this would be too much of a change of the basic flow-monitor concept?

 

Best Regards

 

Von: Gustavo Carneiro [mailto:gjcarneiro at gmail.com] 
Gesendet: Dienstag, 14. Februar 2012 16:57
An: Sebastian Rohde
Cc: ns-developers at isi.edu
Betreff: Re: [Ns-developers] flow-monitor features: time variant scenarios

 

On Tue, Feb 14, 2012 at 15:43, Sebastian Rohde < <mailto:sebastian.rohde at tu-dortmund.de> sebastian.rohde at tu-dortmund.de> wrote:

Dear Gustavo,

Dear all,



we just looked into the flow-monitor and we saw that there is no way of
analysing time variant scenarios. We very often analyse behaviour (i.e.
throughput, delay, jitter on the application layer) of communication
protocols on the time axis. This is particularly useful when looking into
pre-planned or protocol-dependent mobility patterns.



We missed this feature in the flow monitor, but we like the flow-monitor
concept and general architecture very much. Is there anyone currently
working on improving the flow-monitor in this direction? Otherwise we would
go into this direction ourselves.

 

I think there have been people with the same problem in the past, and they probably solved it, but didn't submit patches.  I am not aware of work underway to do this.

 

There are two possible approaches to this problem (maybe others I don't see):

 

1. You use Flow Monitor as is, and periodically call the Serialize method; later, the post-processing stage will probably need to do some more computation to subtract the statistics of one period from the statistics of the period before, to get an idea of what changed;

 

2. You do as above, but also add a new method to Flow Monitor to clear the statistics after serializing them; this way, the post processing stage has less work to do.

 

Best regards.

 



Best regards,

Sebastian



______________________________________________

Sebastian Rohde, M.Sc.

Communication Networks Institute (CNI)

Dortmund University of Technology



E-mail: sebastian.rohde at tu-dortmund.de

http://www.kn.e-technik.tu-dortmund.de/








 

-- 
Gustavo J. A. M. Carneiro
INESC Porto, UTM, WiN, http://win.inescporto.pt/gjc
"The universe is always one step beyond logic." -- Frank Herbert