[tor-relays] Qualities of a good relay (Sean Saito)

Roger Dingledine arma at mit.edu
Wed Jun 24 12:20:26 UTC 2015


On Wed, Jun 24, 2015 at 01:07:28AM +0000, saitosean at ymail.com wrote:
> Besides the obvious requirements of a good relay (e.g. speed,
>geo-diversity, constant uptime), what qualities make a relay valuable
>to the Tor network and its users?

As mentioned in the other replies, "consistently runs a recent enough
version of Tor" is a great one.

Something about exit policy also seems wise.

I wonder if having a ContactInfo counts? I certainly have a little moment
of dread when I run across a relay that doesn't have its Contactinfo set.

Also, OS diversity might be a nice quality to look at.

You might find Mike's Torflow paper interesting here:
https://blog.torproject.org/blog/torflow-node-capacity-integrity-and-reliability-measurements-hotpets
In particular, a relay that doesn't have permission from its OS to use
enough file descriptors is going to be bad news. Similarly, a relay is
less good if it rate limits to the point that it often has cells queued
up waiting for the next interval to arrive (so that's not just declared
bandwidth, but actual resulting performance in practice). Those are both
the sort of thing that need active probing though.

--Roger



More information about the tor-relays mailing list