[tor-relays] bandwidth authority algorithm is cracked

[Roger Dingledine]

On Wed, Jan 22, 2014 at 01:02:29PM -0500, starlight.2014q1 at binnacle.cx wrote:
> >Also keep in mind that what the bandwidth
> >authorities actually measure is not total capacity
> >but spare stream capacity (by downloading large
> >files through at least 5 different two hop
> >circuits times for each relay).
> 
> Wait...  So if I understand this correctly
> the bandwidth number is the difference between
> the actual bandwidth of the node and the
> actively utilized bandwidth?

No. The bandwidth number is how much attention clients should give the
relay. That is, it's a weighting that clients use when selecting relays,
and they select a given relay proportional to its bandwidth weight.

> I looked high-and-low for a proper definition
> of the value

This is where the consensus weight is described:
https://gitweb.torproject.org/torspec.git/blob/HEAD:/dir-spec.txt#l1634

The consensus weight is computed using
a) the relay's self-advertised bandwidth in its descriptor:
https://gitweb.torproject.org/torspec.git/blob/HEAD:/dir-spec.txt#l389
b) the ratios of bandwidth weights for various types of relays:
https://gitweb.torproject.org/torspec.git/blob/HEAD:/dir-spec.txt#l2137
and
c) the result of the active measurements from the bandwidth authorities:
https://gitweb.torproject.org/torflow.git/blob/HEAD:/NetworkScanners/BwAuthority/README.BwAuthorities
https://gitweb.torproject.org/torflow.git/blob/HEAD:/NetworkScanners/BwAuthority/README.spec.txt

> However I must complain that the consensus
> bandwidth does not say much about the 
> relative health of a relay.  Does a good
> (ungamable) way exist to show the bandwidth
> capacity of relays along with the available
> capacity?

Ungameable is a much harder requirement. See e.g.
http://freehaven.net/anonbib/#eigenspeed
but Eigenspeed does not do well at assessing the speed of fast relays,
since it's hard to design a system where relays can accurately assess
the speed of faster relays.

--Roger