Am Samstag, 9. Februar 2013 schrieb grarpamp : > > I'm planning to do > > the same thing for measuring hidden service performance in Torperf. > > > there's a quite high probability that > > you're using the same entry guard for the actual connection between > > client and service. > > For similar HS tests I ran the HS's separately so that, other than the > local ISP, it would best reflect the actual Tor network. > Though doing it all with one Tor instance would be much easier. > As time frees up I may continue and contribute to your project. > > In addition to the EG, I wonder as to possible reuse of the paths to the > DB's, > IP's, and RP? It wasn't something I looked at, but could also affect > performance.
You have a point here. My plan for these measurements was to disable guards which I do for all measurements. But using a single Tor process for both client and service means that it will sooner run out of pre-built circuits. So maybe it's more realistic to use two separate Tor processes. This shouldn't be hard with the redesigned Torperf. By the way, if you're interested in the current state of the rewrite, see the perfd branch in https://gitweb.torproject.org/karsten/torperf.git . Still far from being complete, but on the right path, I hope. Contributions would be cool! Best, Karsten _______________________________________________ tor-talk mailing list tor-talk@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-talk
