[tor-talk] Tor network too fast
arma at mit.edu
Thu Dec 11 03:25:49 UTC 2014
On Wed, Dec 10, 2014 at 10:05:36PM -0500, Griffin Boyce wrote:
> Sieme wrote:
> >I think so too. But what's the default policy now? 3 hops?
> It's three hops. 100ms is pretty short, but not ridiculously so.
In particular, a lot of the slowness of Tor before (and still, if
you get unlucky) is because of queueing inside the relays. Tor isn't,
and never was, slow because of network latency. It's slow (when it is)
because other people are using it at the same time, and their traffic
is going through before yours so yours has to wait its turn. And it gets
especially bad when the relay you and the other person have selected is
rate limiting such that the relay stops and sits on your traffic data
until it's willing to send some more of it.
That's why designs like KIST can show such dramatic improvements (in the
lab, but I'm hopeful they will show improvements in reality too)
because they do a better job of prioritizing the cells that will
actually get through, rather than choosing ones that will just end up
More information about the tor-talk