> -----Original Message-----
> From: [EMAIL PROTECTED] 
> [mailto:[EMAIL PROTECTED] On Behalf Of Martin 
> Stone Davis
> Sent: den 19 november 2003 10:08
> To: [EMAIL PROTECTED]
> Subject: [freenet-dev] Re: High 'Data waiting to be transfered'
> 
> 
> Toad wrote:
> 
> > On Tue, Nov 18, 2003 at 03:46:19PM -0800, Martin Stone Davis wrote:
> > 
> >>Toad wrote:
> >>
> >>
> >>>On Tue, Nov 18, 2003 at 11:00:27PM +0100, Thomas Themel wrote:
> >>>
> >>>
> >>>>Hi,
> >>>>
> >>>>my node is currently running off an ISDN link, and it 
> seems to build 
> >>>>up
> >>>>pretty high values in the connection manager's 'Data waiting to be
> >>>>transfered' counter. It's constantly around 30MiB, and 
> that's more than 
> >>>>the
> >>>>link can do in an hour under optimal conditions - I don't 
> believe this
> >>>>makes sense. Or does it?
> >>>>
> >>>>Is there any way to get more sensible numbers?
> >>>>
> >>>>I've tried opening a lot of freesites, and they always seem to 
> >>>>restart. However, some give me the "The request couldn't 
> even make 
> >>>>it off of your node." message, on a node that wasn't reseeded for 
> >>>>months.
> >>>>
> >>>>Is this related?
> >>>>
> >>>>I get the feeling that the behaviour has worsened in recent 
> >>>>versions, it used to work nicely around 632x.
> >>>
> >>>
> >>>I suspect this is because we reinstated the QR based on bandwidth 
> >>>code.
> >>
> >>Can you explain how that follows?  I remember that before that code 
> >>was
> >>reinstated, I had 300-400 transmitting connections, and my 
> output bw of 
> >>trailers was reduced down to about 50% of total.  Now I 
> trasmit on about 
> >>30 connections and the trailer bw is about 90% of total.
> > 
> > 
> > No. Bandwidth, # transfers, is better. Routing is worse.
> 
> By what measure is routing worse?
> 
> Can you show how removing bandwidth-based-QR would improve "routing" 
> without harming the network as a whole by reducing bandwidth and # of 
> transfers?

NGR is designed to take into account a nodes available bw and include
that fact in routing decisions. Instead of overloading an already
overloaded (bw-wise) node even more NGR would be sending the Q to
another node.

A bw-caused QR will completely hide the actual bw problem from other
nodes' NGRT:s since the 'load' that is actually causing a QR includes
multiple other factors.


/N

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

Reply via email to