Thus spake Jim (Jimmymac@xxxxxxxxxx): > Olaf Selke wrote: > >On 06.09.2010 19:25, Moritz Bartl wrote: > >>On 06.09.2010 18:07, Olaf Selke wrote: > >>>depending on your tcp timeout parameters you need a lot of local ports. > >>>Did you verify the number of Tor tcp sessions isn't limited by > >>>net.ipv4.ip_local_port_range? > >>Yes. > > > >I'm clueless. Do you want me to shut down all four blutmagie exits for a > >couple of days and wait if other exits get more traffic? > > If the idea is to run some experiments, I have a proposal that perhaps > would have less adverse impact on the Tor network -- and hoperfully > still yield useful data ... > > I believe torservers (Moritz's nodes) are runnning slighly less than 3/4 > of what he believes their capacity should be. And I think he is > running four instances of roughtly equal capacity. What if Moritz -- as > an experiment -- converted one of his four nodes to a non-exit node. > If his total traffic is roughly unchanged from what it is now, then I > believe that would indicate (relative) excess exit capacity is not the > problem. If his total traffic increases significantly, then perhaps > further investigation is warranted. Is that worth a try? The way to do this is to just spin up additional nodes. I believe Moritz has tried this, and so has Jake. AFAIK, it did help Jake hit 1Gbit for Guard+Middle traffic. I'm not sure on Moritz's results, though. -- Mike Perry Mad Computer Scientist fscked.org evil labs
Attachment:
pgpq2XzD8gJqn.pgp
Description: PGP signature