Aquatorius wrote: > > I've "only" a 10 MBit connection but I do see this behaviour too (Tor > 0.1.2.17 at the moment): > > http://aquatorius.spemaus.net/munin/spemaus.net/aquatorius.spemaus.net.html#Network > > I usually restart Tor every 5 days when network utilization decreases due > to heavy swapping, see memory usage graph. But this strange traffic pattern > repeats almost daily.
my tor's cpu load and number of concurrent tcp connection graphs do confirm Aquatorius' readings. (The cpu load graph may exceed 100% due to four cores - 2xXeon hyper threading cpu). In comparison with tor the extra traffic and cpu load caused by the Apache web server http://torstatus.blutmagie.de running on the same box is peanuts. btw: besides the German BKA (German 'FBI') the German Verfassungsschutz (Federal Office for the Protection of the Constitution) does know about tor, and they feel quite uncomfortable with it. It might not be desirable for the tor network if a few servers do exit most of the traffic. On the other hand pushing the limits always is challenging. I would be happy if tor needs less memory and improves support of multiple cores/cpus. Even with numcpu parameter set to four, one core is running the main tor thread with max load and the other three are more or less idle (load <10%) running the helper threads. regards, Olaf
<<inline: anonymizer.blutmagie.de.tcp-week.png>>
<<inline: anonymizer.blutmagie.de.cpusum-week.png>>
