On Sat, Nov 22, 2003 at 06:22:57PM +0200, Jusa Saari wrote:
> On Fri, 21 Nov 2003 22:47:21 +0000, Toad wrote:
> 
> > On Fri, Nov 21, 2003 at 03:20:58PM -0500, Ed Tomlinson wrote:
> >> On November 21, 2003 01:45 pm, Toad wrote:
> >> > On Fri, Nov 21, 2003 at 12:01:27PM -0500, Ed Tomlinson wrote:
> >> > > On November 21, 2003 06:41 am, Ian Clarke wrote:
> >> > > > It seems that we aren't seeing the hoped-for specialization in NGR.
> >> > > > Rather than futzing about with all sorts of ideas while trying to test
> >> > > > them in the chaotic real network, we need a simple simulator to help us
> >> > > > test these ideas.
> >> > >
> >> > > One idea to test on that simulator is are we specializing by key or
> >> > > for fast connections.  It seems to me that, by far, the largest chunks
> >> > > of the NG estimator are from transmission time...  When I did the
> >> > > first stabs at NG I was only basing it on search times and, if
> >> > > memory serves, it was getting much better numbers.
> >> > >
> >> > > Maybe we should be normallizing to a size that does not give
> >> > > a number much greater than transmission time?
> >> >
> >> > Why?
> >> 
> >> The search time is usually a number averaging around 15,000 ms
> >> The time to recieve a normalized file is over 150,000 ms.  This
> >> means the 15000 (or 2000 or 30000) gets lost in the noise and NGR 
> >> just looks for the fastest node.
> > 
> > Those are not the only factors involved. The transfer time is multiplied
> > by the probability of a transfer occurring.
> 
> 
> I must ask... Doesn't this mean that the worse the node's chances of
> finding a given key, the better (lower) the estimate ? Shouldn't you
> multiply by the (1 - probability of transfer occurring) ?

No. The possibility of failure is also taken into account.

Further, a node that is better at a given key subspace should also transfer
data faster in that area.
> 
> Unfortunately, I was unable to find the formula of estimate in the CVS
> source, so I couldn't check... Anyway, if this is true, it would explain
> the despecialization.

StandardNodeEstimator.longEstimate().
-- 
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