[tor-relays] Advantage in more exits in the same /8?

[Moritz Bartl]

Hi Jesse,

On 08/26/2014 03:47 PM, Jesse Victors wrote:
> It seems to me that too many nodes under the same ISP is 
> problematic because it concentrates too much traffic in
> the same AS, but on the other hand, Tor could use more exits.
> More importantly, how many is too many nodes in the same /8,
> or in the same /16? Where would you draw the line?

Very good question. Ideally, the Tor client would be AS-aware, and you
would not have to worry about it. For the interested reader, see for
example [1]. Until then, my thinking is that I compare to other
locations. https://compass.torproject.org/ is very helpful for that: For
example, if you group by AS, the largest AS right now (i3d, NL) in
regards to exit capacity has 11%, and OVH tops the overall network at
10% consensus weight.

As a rough rule, I'd avoid to push more than 1-2Gbit/s of traffic at one
ISP. On the plus side, as long as you don't top the list, you're
weighing down other locations. And universities are a preferred location.

Make sure to use the MyFamily statement correctly: Unless relays are on
the same /16, Tor might pick multiple of them for a circuit. Also, if
you want to push more than ~100 Mbit/s on a single machine, you need
AES-NI or run multiple relays, for more than 400 Mbit/s you need to run
multiple relays in any case. The multi-relay initscript can be quite
helpful for that.

[1] http://freehaven.net/anonbib/#ccs2013-usersrouted
[2] https://www.torservers.net/wiki/setup/server#multiple_tor_processes

-- 
Moritz Bartl
https://www.torservers.net/