On November 07, 2003 10:29 pm, Zlatin Balevsky wrote: > >One radical solution: > >Remove the code to reject queries when the bandwidth limit is exceeded! > > > which returns us in the state 5010-5018 where the node has accepted > umpteen transfers, each going at snail speed. Making that prevalent > accross the network will be catastrophic.
I see that now, with just the probablistic rejection. > Another possibility is to go even further back in 5007-5010 where the > node would still accept everything but wouldn't round-robin to the > senders which effectively meant the queries were served in a fifo queue. What I have tried previously is to queue trailers (in PeerHandler) and not start sending them until we have bandwidth. To make this work I had to setup and varient of decayingRunningAverage that used a 5-10 second window to predict bandwidth. Using the current bandwidth numbers, based on a 1 minute window, lead to 'deadtime' where we did not sent at all... With the above code it place I would normally have 8-15 transmitting connections and a queue of 1-4M. Which seems much more reasonable given that I have use a 10K limit. Ed _______________________________________________ Devl mailing list [EMAIL PROTECTED] http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl
