[Simgrid-user] network model

Arnaud Legrand arnaud.legrand at imag.fr
Tue Sep 13 11:14:05 CEST 2011


Hi,

Le dimanche 11 septembre 2011 à 12:54 +0100, karim idres a écrit :

> i juste wont to ask another question relating to network models .now
> i'am writing a chapter about simgrid ,i realy need to have  more
> details on network/cpu/worksation models 
> how this models work ,there formulas and meaning .
> iknew LV08,CM02 but nothing aboute Reno ,Reno2,Vivaldi .

The most up-to-date article about LV08 and CM2 is here
(http://simgrid.gforge.inria.fr/publis_core.html):

        Pedro Velho and Arnaud Legrand. Accuracy Study and Improvement
        of Network Simulation in the SimGrid Framework. In Proceedings
        of the 2nd International Conference on Simulation Tools and
        Techniques (SIMUTools'09), Rome, Italy, March 2009.
        
To know more about the list of available models, check this:

  http://simgrid.gforge.inria.fr/3.6.1/doc/options.html

Reno, Reno2 and Vegas are implementations of the fluid models of TCP
Reno, NewReno and Vegas proposed in
        S. H. Low, “A duality model of TCP and queue management
        algorithms,” IEEE/ACM Transactions on Networking,
        vol. 11, no. 4, 2003.
They are based on non-linear (but hopefully convex) optimization
programs and are implemented in a rather efficient way in SimGrid. Yet,
our findings are that, although these models come from a very
interesting low-level analysis of the protocol behavior and should be
much more sound than the LV08 and CM02 models, they are unable to
account for some cross-trafic interferences. We are currently writing
the article on this finding and will submit it to publication in the
next months. We will keep you posted on the outcome. This work was done
in Pedro Velho's PhD thesis (http://www.liglab.fr/spip.php?article1013),
which you can reference as follows:

        @PhdThesis{pedro_phd,
          author = 	 {Pedro Velho},
          title = 	 {Accurate and Fast Simulations of Large- Scale Distributed Computing Systems},
          school = 	 {University Joseph Fourier},
          year = 	 {2011},
          address = 	 {Grenoble},
          month = 	 jun,
        }

Best regards,

    Arnaud

-- 
Premature optimization is  the root of all evil.
               -- D.E.Knuth




More information about the Simgrid-user mailing list