[tor-relays] bwadv vs bwhist + implications

[torhoste at tor.host-ed.me]

Hi,

we're currently looking forward to expand our contribution to the Tor
network and as a mathematically fascinated person, I'm pretty much into
statistical data from https://metrics.torproject.org/network.html at the
moment and implication on Accounting/Bandwidthrate configuration
opportunities for our Tor relays to utilize them at maximum efficiency,
i.e. I'd like to start a discussion with other relay operators.

That said, I noticed that the BW advertised into the network (bwadv) vs
the BW actually put through the network (bwhist) is at a 1.544945508 ratio
throughout the whole Tor network. I.e. a relay configured to provide 500
GiB of traffic/mo would put through (respectively actually is putting
through right now) "just" 323.63601 GiB traffic/mo on average.

Now, my questions (respectively thoughts for discussion) would be:

1. What would you assume to be the reason for it? I mean this is a
calculated average through 3117 relays and thus I'd say rather meaningful,
i.e. the largest possible "sample" (i.e. it isn't just a few relays
behaving like so). Also, looking at one of our own two test relays that is
*definitely* running 24/7 - vks24949, it pretty much shows the same
figures. Vnstat estimates 342.23 GiB traffic/mo for September 12 and the
relay actually is configured to provide 500 GiB traffic/mo with daily
accounting set up (1.461005756 ratio vs 1.544945508 ratio calculated from
*.csv provided at https://metrics.torproject.org/network.html for the
whole Tor network).

2. As a result of all that, I currently come to the conclusion that it
could be a good practice to actually configure 1.544945508 times the
traffic for both the daily accountingmax + the actual relaybandwidthrate
in KiB/s to *really* utilize the BW provided by sponsors (in our very
case) to the full potential.

I'm looking forward to your input on this. Thanks in advance,
Thomas