On Mon, Oct 27, 2003 at 07:01:51AM -0500, Ed Tomlinson wrote:
> On October 26, 2003 09:54 pm, Ken Corson wrote:
> > Okay, this one took me 20 minutes to catch it happening - my node is
> > rejecting the majority of query requests. And it works like this:
> >
> > I accept every request for a 'little while' (1 to 2 minutes). Then I
> > reject every request (for 5 to 20 minutes) until it is time to accept
> > them all again. And the cycle repeats.
> >    I noticed a pattern like this in localQueryTraffic by minute :
> >
> > 10/26/03 9:24:00 PM EST     96      0       0.0
> > 10/26/03 9:25:00 PM EST     132     0       0.0
> > 10/26/03 9:26:00 PM EST     189     26      0.13756613756613756
> > 10/26/03 9:27:00 PM EST     595     463     0.7781512605042017
> > 10/26/03 9:28:00 PM EST     531     0       0.0
> > 10/26/03 9:29:00 PM EST     323     0       0.0
> > 10/26/03 9:30:00 PM EST     556     0       0.0
> > 10/26/03 9:31:00 PM EST     438     0       0.0
> > 10/26/03 9:32:00 PM EST     401     216     0.5386533665835411
> > 10/26/03 9:33:00 PM EST     710     472     0.6647887323943662
> > 10/26/03 9:34:00 PM EST     367     0       0.0
> > 10/26/03 9:35:00 PM EST     250     0       0.0
> > 10/26/03 9:36:00 PM EST     123     0       0.0
> > 10/26/03 9:37:00 PM EST     308     0       0.0
> >
> > the ratio is usually 0. So, I wanted to know, if during the
> > active periods, there were ANY rejects. The way I determined that
> > there were NOT, was by repeatedly clicking on  Network Load .
> 
> I have also noticed this.  Its been happening for many builds.  There 
> were some patches posted a while back to do probalistic backoff.  
> Maybe these would smooth things out?
> 
> The other thing this sort of usage pattern does is make other nodes
> see higher values for SendFailed than are real.  The DecayingRunningAverage
> class is very quick to move to 1.0...  I do have a variant on this class
> that uses a window of several minutes to determine the number to
> feed the running average with.  It seems to get more reasonable 
> numbers for SendFailed.

Is that bad though? Maybe we want it to be pretty sensitive at least for
sendfailed, so it backs off the node while it is rejecting queries?
> 
> Ed

-- 
Matthew J Toseland - [EMAIL PROTECTED]
Freenet Project Official Codemonkey - http://freenetproject.org/
ICTHUS - Nothing is impossible. Our Boss says so.

Attachment: signature.asc
Description: Digital signature

_______________________________________________
Devl mailing list
[EMAIL PROTECTED]
http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl

Reply via email to