Ken Corson wrote:
Niklas Bergh wrote:

The real question is: how much of your outgoing bw
is used up by QR:s, how much is used by routing requests, and how much
is used up by "real" data.  I wish I knew how to tell this.


absolutely! Hopefully we agree that minimizing the percent of
bandwidth spent on QR's is a desirable goal. It appears that my
outbound bandwidth is divided "about" 65/35 trailers:QRs, and
that's not a good thing.
Perhaps, but it would be nice to get a more datailed and accurate assessment. We don't know *for sure* what that 65% and 35% really are... As Iakin says, some of that 35% might be "real" data (trailers); also, some of it may be routing requests.


> But it's comforting to have a sense of
the "waste" , because any QR is a waste of network bytes. It does
provide a little bit of info (that I'm busy) to the requestor.
It certainly isn't an efficient way to inform though. My concern
is the manner in which my node accepts/rejects. For 10 minutes,
the window is closed. Then for 48 seconds, the window is wide
open, and all requests are accepted. Then the window slams shut
again.
Same thing here.

What this would mean for most peers, IF they queried at
a constant rate, is that only 1 of 10 get through to me, and
they must pick a different route most of the time. Yet for
some reason they chose to route to me (first? i dunno!).

Most nodes will QR most of the time, so the fact that yours does too shouldn't scare them off. And it makes sense to query your node that will probably QR because at least you get the failure fast. What really matters is how long to get a success.

  If 80% of queries are being rejected, globally, this implies
that there are 'too many' requests being made, and _we_are_
_attempting_to_throttle_them_ , albeit passively by a receiver,
rather than a sender. Long-term - what if sender and receiver
could 'tune' an optimal rate ? Ian very recently suggested a
backoff deal, like "squelch the next 5" or something similar.


[snipped ... moved to "Discrepancy between reported and actual output bytes" thread]


Thank you both for educating me! Now I know a little more - and I can better complain ;)

BTW Neither of you answered the original question - what are
your rejection rates ? I'm not convinced that they just don't
matter.


From /servlet/nodestatus/diagnostics/localQueryTraffic/hour, I see that I reject about 95% of all queries.


-Martin


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

Reply via email to