[tor-relays] Home broadband - worth running a relay?
arma at mit.edu
Sat Jul 13 21:39:01 UTC 2013
On Thu, Jul 11, 2013 at 09:43:00PM +0100, Nick wrote:
> I have a reasonable ADSL connection, and a little always-on server.
> The bandwidth is in the region of 2Mib/s down, something less up
> (maybe 256Kib/s). Is it useful for me to run a tor relay with this
> bandwidth? I'd like to run one which isn't an exit, at least for
I think at this point if you're at least 800kbit (100KBytes/s) each way,
it's useful to be a relay.
Here are the flags thresholds that moria1 (my directory authority)
is voting for right now:
flag-thresholds stable-uptime=646220 stable-mtbf=1055595 fast-speed=18000
guard-wfu=98.000% guard-tk=691200 guard-bw-inc-exits=185000
guard-bw-exc-exits=148000 enough-mtbf=1 ignoring-advertised-bws=1
So that means if you have 18KBytes/s you get some use, and if you have
either 185KBytes/s or 148KBytes/s (depending on if there's enough exit
capacity) you can get the Guard flag if you're stable enough:
It's been suggested by some researchers that a wiser lower bound on
whether a relay is useful is 8Mbit/s each way:
But I think we'd be throwing out enough capacity at that point that it's
not clear to me that the performance gains would be seen in practice.
Plus there's some impact to relay diversity (aka anonymity). Not to
mention the impact to the community when you tell well over half the
relays that actually you don't need them thanks bye.
> If not, am I correct in thinking that a bridge is an appropriate
> help? That's what I'm doing currently, but if a relay would be more
> useful I'd be very happy to do that.
Yes, a bridge is a fine thing to run on a connection with 250KBytes down
and 32KBytes up.
In the future, we might end up with a system like Conflux to let you
glue together two slow bridges and get better throughput:
> One other unrelated(ish) question: I'm in the UK, where the idea of
> censorship isn't resisted as strongly as it ought to be, and as a
> result my internet connection is subject to a smallish amount of
> censorship: whatever is on the secret IWF blacklist plus the pirate
> bay. Does this mean that running an exit node from a home connection
> here at some point in the future would not be helpful? Or only if
> all HTTP(S) was blocked (as the IWF blacklist is secret there's
> presumably no way to tell the tor network what is inaccessible from
> this node).
This is a tough one. It's not like the web is divided into the clean
uncensored part and the dirty censored part -- it's all getting to be
grey. If we demanded all exits be totally uncensored, we'd quickly run
out of qualified places for exits. In the ideal case, the censorship
at your exit would be in the form of not knowing the answer to a DNS
resolve of the destination -- and in that case Tor will automatically
fall back to trying a different exit. That still isn't ideal though,
since it would make things slower (since you'd have to time out before
switching to a different circuit), and if it happened a lot then it
could take a long time to find a workable circuit.
More information about the tor-relays