On Sat, Aug 22, 2009 at 2:05 AM, Toni Bergman<toni.berg...@gmail.com> wrote: > Freenet 0.7.5 Build #1232 build01232 > Freenet-ext Build #26 r23771 > > So far I've understood the backoff percentages in advanced connection > details but no longer. > The attached picture shows what I'm saying. > > What ticks me off the most is being connected for over 90 minutes to peers > that are 99,9 % of the time backed off. This defeats the purpose of having > 32 peers with output of 90 kB/s. > The bug I'm reporting is something I didn't see before. > Note the values backed off wait time remaining and time total. Time total is > now smaller than the time remaining. > Seems to be happening with several peers. > > Hows about giving the peers a loser rating and connect the losers together. > Sure they won't be able to download anything but they won't be slowing me > down either, which is the most important thing, since I'm the most important > person in the world.
How much bandwidth usage did you see in practice before, with the 90 KiB/s setting and 20 peers? There's an argument to be made we gave too many peers to slow nodes with the recent changes. Previously, people with fast connections would see 80-100 KiB/s usage (my rough estimate, not looking at actual numbers right now). Now we have 31 peers at 80 KiB/s, despite the fact that it wasn't required before. Which means that the high peer counts aren't helping... I'm not advocating changing anything, and certainly not changing it right away; we need more data first (such as we can get :/ ). It's worth thinking about changing the constants to require more bandwidth per peer, though. Specifically, we should look at the limited data we have, and set the constant such that the (pre-changes) maximum bw usage with 20 peers corresponds to 20 peers or less with the new code. Evan Daniel _______________________________________________ Support mailing list Support@freenetproject.org http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://emu.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:support-requ...@freenetproject.org?subject=unsubscribe
