Each authority singing its own song?

There seems to be a huge discrepancy between what each authority reports as a part of their consensus especially when it comes to bw report which turns out to be what’s reported as consensus weight.

I have 2 Guard relays running on identical VMs. Same OS, same network card config, same torrc (Except for the name). The bw reported by each authority is 22000, 30000, 33000, 39000, 45000 and consensus of 33000

The other one which is identical shows as 20000, 22000, 40000, 41000, 55000 and consensus of 40000.

Both of them had a consensus of 48000 - 53000 just a few weeks ago. Neither of them is overloaded either.

A quick random look at other relays in the consensus, I see a relay with bw reports in a huge range of 20000 to 120000 for the same relay reported by different authorities. Another one 40000 to 110000. So I’m wondering. What gives? what is the basis of these measurements and why can’t authorities reach a similar conclusions?

There’s also inaccuracy in what’s reported as MaxAdvertisedbandwidth. I’m running 4 bridges which are run on cloned VMs that are identical in every way including their OS and torrc and the maxAdvertisedBandwidth for them is 6MiB/s. One shows as 6, the rest show as 5.81, 4.34, 3.93.

2 Guard relays advertising 20 MiB/s, showing as 19.66 and 19.23.

Any ideas?

Authorities checks are random and relay load is random because of random user behaviour.

The five SBWS select different nodes to build 2-hop circuits. Since your node is a guard node, I guess the exit node chosen by SBWS may have adversely affected your node’s bandwidth measurement. Thus, the five SBWS have different bandwidth measurements for two nodes of the same host.

You can compare the measurement of the same node at the current moment with the measurement of the previous hour to see if there is a significant change.