Time Tries Successes Ratio 8/5/03 7:00:00 PM EDT 3961 2230 0.5629891441555163 8/5/03 8:00:00 PM EDT 3757 1895 0.5043918019696566 8/5/03 9:00:00 PM EDT 4066 1619 0.39818002951303494 8/5/03 10:00:00 PM EDT 7583 896 0.11815903995780035 8/5/03 11:00:00 PM EDT 3876 1381 0.3562951496388029 8/6/03 12:00:00 AM EDT 5181 1575 0.3039953676896352 8/6/03 1:00:00 AM EDT 5861 1299 0.22163453335608257 8/6/03 2:00:00 AM EDT 7631 2507 0.3285283711178089 8/6/03 3:00:00 AM EDT 8932 2305 0.25806090461262876 8/6/03 4:00:00 AM EDT 9569 1642 0.1715957780332323 8/6/03 5:00:00 AM EDT 10812 1682 0.15556788753237144 8/6/03 6:00:00 AM EDT 11816 1074 0.09089370345294516 8/6/03 7:00:00 AM EDT 8790 1272 0.1447098976109215 8/6/03 8:00:00 AM EDT 11346 1181 0.10408954697690816 8/6/03 9:00:00 AM EDT 9999 756 0.07560756075607561 8/6/03 10:00:00 AM EDT 8622 1046 0.1213175597309209 8/6/03 11:00:00 AM EDT 7570 981 0.1295904887714663 8/6/03 12:00:00 PM EDT 6035 733 0.12145816072908036 8/6/03 1:00:00 PM EDT 7207 1272 0.1764950742333842 8/6/03 2:00:00 PM EDT 6345 795 0.12529550827423167 8/6/03 3:00:00 PM EDT 6710 1156 0.1722801788375559 8/6/03 4:00:00 PM EDT 6022 1071 0.177847891066091
** At this point, I changed maximumThreads from the default (120?) to 200, and then restarted. ** 8/6/03 5:00:00 PM EDT 5414 3813 0.7042851865533801 8/6/03 6:00:00 PM EDT 6030 4325 0.7172470978441128 8/6/03 7:00:00 PM EDT 6533 4418 0.6762589928057554 I'd say that's a pretty dramatic change. Is there any good guideline for how high maximumThreads should be set? Or is this something that node operators are going to have to adjust themselves, using nothing more than simple trial and error, to achieve the best results? This is build 5021, on Linux 2.4.21 (non-NPTL), with Sun 1.4.1_03-b02. I am limited by bandwidth, not CPU; this is running on an Athlon XP 2000+ (1.67 GHz), but my ADSL has an upstream of 256kbps. I've got the Freenet outputBandwidthLimit set to 16000, but I might be able to increase that a bit. I'm also doing queueing on the router (my OpenBSD box), with Freenet packets getting a priority just under the default (but higher than giFT). -- Greg Wooledge | "Truth belongs to everybody." [EMAIL PROTECTED] | - The Red Hot Chili Peppers http://wooledge.org/~greg/ |
pgp00000.pgp
Description: PGP signature
