[ns] Bug in NS-2.1b8a for ECN with "gentle_" ???

Lisa Wu c16917@yahoo.com
Mon Apr 29 12:20:03 2002


Hi,

I have one similar problem as listed below. It is
supposed the marking (dropping) probability should
vary btw 10% and 100% with the parameter gentle_
setting.

The program does calculate the different probability
for the situation avg_<th_max and
th_max<avg_<2*th_max.
However,in the function drop_early(pkt), the condition
to set Congestion Experience bit is
 
if(edp.setbit && hf->ect() && edv.v_ave<edp_.th_max){
	hf->ce()=1;
	return (0);
}

It only considers the case in which the parameter
gentle_ do not set. But when has the gentle_, it means
under certain probability (u<=edv_.v_prob), the packet
is dropped instead of marking when avg>th_max. I think
it should add another judgement, such as,

if(edp.setbit && hf->ect() && edv.v_ave<edp_.th_max &&
!gentle_){
	hf->ce()=1;
	return (0);
}
else if(edp.setbit && hf->ect() && gentle &&
edv.v_ave<2*edp_.th_max){
 	hf->ce()=1;
	return (0);
}
....

I am not sure my understanding is right or not. And I
am looking forward any progress on this issue. Thank
you.

Lisa
 
-- Frank Akujobi <bulggie@yahoo.com> wrote:
> 
> Hi,
> Thanks for your prompt response. The problem is
> thus:
> Within the region of maxth_ < avg_ < 2*maxth_ and
> with
> "gentle_" parameter enabled, the drop probability is
> supposed to vary btw. 10% and 100%. When a drop
> occurs
> under this situation it is supposed to be a FORCED
> drop. 
> In the codes however (function drop_early() in
> red.cc), we see that for ECN the marking
> probability(which should be the drop prob. for RED)
> is
> used only within the region of minth_ < avg_ <
> maxth_.
> If avg_ happens to be above maxth_ but less than
> 2*maxth_ packets are dropped with a probability of
> 100% (as UNFORCED drop) NOT a range btw. 10% and
> 100%
> as required. This is the problem I see. I am hoping
> that you can confirm this or probably show me what I
> don't understand. 
> 
> Hope this helps and looking forward to hearing from
> you.  
> 
> Frank Akujobi.
> 
> 
> 
> 
> 
>  --- Xuan Chen <xuanc@isi.edu> wrote: > I will try
> to
> find a right person for you
> > (hopefully..:)). but, in 
> > the mean time, could you  please describe the
> > problem (original code, your 
> > fix, and the simulation results) in details for
> > ns-users?
> > 
> > Thanks,
> > -chen
> > 
> > On Thu, 28 Feb 2002, Frank Akujobi wrote:
> > 
> > > Hi,
> > > I am working on some form of ECN and I observe
> > > difference in results when I provide a fix. I
> had
> > > thought there was a reason for writing the codes
> > like
> > > that, otherwise it could be a bug.
> > > If U are not familiar with the area, pls can you
> > refer
> > > to someone amongst your NS "administrators" who
> > might
> > > be able to clear up any doubts we have.
> > > 
> > > Thanks and hoping to hear from U.
> > > Frank.
> > > 
> > > 
> > > 
> > >  --- Xuan Chen <xuanc@isi.edu> wrote: > I am
> sorry
> > > that I am not familiar with that part of
> > > > the code. Have you got 
> > > > any difference in simulation results with the
> > > > original implementation and 
> > > > your fix?
> > > > 
> > > > -chen
> > > > 
> > > >  On Wed, 27 Feb 2002, Frank Akujobi wrote:
> > > > 
> > > > > Hi,
> > > > > I had posted this on the mailing list 2 days
> > ago
> > > > but
> > > > > till now there's been no reponse. Pls could
> > you
> > > > find
> > > > > time to consider my comments......
> > > > > 
> > > > > With ECN and "gentle_" parameter set, packet
> > > > marking
> > > > > prob. should vary 
> > > > > betw. 10% 
> > > > > and 100% when  maxth_ < avg_ < 2*maxth_. In
> > the
> > > > > NS-2.1b8a, I find the 
> > > > > following 
> > > > > codes in the function drop_early() within
> > red.cc.
> > > > > 
> > > > > if (edp_.setbit && hf->ect() && edv_.v_ave <
> > > > > edp_.th_max) {
> > > > > 			hf->ce() = 1; 	// mark Congestion
> > Experienced
> > > > bit
> > > > > 			return (0);	// no drop
> > > > > 		} else {
> > > > > 			return (1);	// drop
> > > > > 		}
> > > > > 
> > > > > This code follows the computation of marking
> > > > > probability for ECN. If 
> > > > > the avg_ is 
> > > > > between maxth_ and 2*maxth_ and gentle_ is
> > > > enabled,
> > > > > the codes above 
> > > > > would always 
> > > > > cause the arriving packet to experience
> > UNFORCED
> > > > drop
> > > > > irrespective of 
> > > > > the 
> > > > > computed probabilty.
> > > > > Is this a bug or is there something I am
> > missing
> > > > out.
> > > > > 
> > > > > Thanks
> > > > > Frank.
> > > > > 
> > > > > 
> > > > > =====
> > > > > Frank Onyekachi Akujobi
> > > > > Broadband Networks Laboratory 
> > > > > Systems and Computer Engineering
> > > > > Carleton University
> > > > > 1125 Colonel By Drive, Ottawa, Ontario
> > > > > K1S 5B6, Canada
> > > > > Cell:(613)286-3738
> > > > > 
> > > > >
> > __________________________________________________
> > > > > Do You Yahoo!?
> > > > > Everything you'll ever need on one web page
> > > > > from News and Sport to Email and Music
> Charts
> > > > > http://uk.my.yahoo.com
> > > > > 
> > > > 
> > > > -- 
> > > > Xuan Chen
> > > > USC/ISI
> > > >  
> > > 
> > > =====
> > > Frank Onyekachi Akujobi
> > > Broadband Networks Laboratory 
> > > Systems and Computer Engineering
> > > Carleton University
> > > 1125 Colonel By Drive, Ottawa, Ontario
> > > K1S 5B6, Canada
> > > Cell:(613)286-3738
> > > 
> > >
> __________________________________________________
> > > Do You Yahoo!?
> > > Everything you'll ever need on one web page
> > > from News and Sport to Email and Music Charts
> > > http://uk.my.yahoo.com
> > > 
> > 
> > -- 
> > Xuan Chen
> > USC/ISI
> >  
> 
> =====
> Frank Onyekachi Akujobi
> Broadband Networks Laboratory 
> Systems and Computer Engineering
> Carleton University
> 1125 Colonel By Drive, Ottawa, Ontario
> K1S 5B6, Canada
> Cell:(613)286-3738
> 
> __________________________________________________
> Do You Yahoo!?
> Everything you'll ever need on one web page
> from News and Sport to Email and Music Charts
> http://uk.my.yahoo.com
> 





__________________________________________________
Do You Yahoo!?
Yahoo! Health - your guide to health and wellness
http://health.yahoo.com