Re: [tor-relays] Drop in consensus weight

[teor <teor2345@xxxxxxxxx>]

(It really would help if you posted your replies below the previous reply.)

> On 9 Nov. 2016, at 10:07, r1610091651 <r1610091651@xxxxxxxxxx> wrote:
> 
> Based on found numbers, some thoughts:
> * shouldn't there be more authority servers? DNS system has 13 root servers spread over the globe…

There are 5 bandwidth authorities, we're working on getting more.

> * not all authority servers are equal:
> ** reported data varies greatly between servers
> https://consensus-health.torproject.org/consensus-health-2016-11-08-19-00.html#36EE8D47E570B8D5515460A9972F3CFD9EDFDFCE: maatuska=1400 <> gabelmoo=4060

Yes, the speed measured by a bandwidth authority depends on:
* location of the bandwidth authority,
* time of measurement,
* randomly chosen relay pair,
* latency and bandwidth through that pair to the remote download site,
* other load on the bandwidth authority and relays at the time.

In general, the further away your relay is from the bandwidth authority, the
lower the measurement will be.

> ** some authorities vary greatly in reported data:
> https://consensus-health.torproject.org/consensus-health-2016-11-08-19-00.html#36EE8D47E570B8D5515460A9972F3CFD9EDFDFCE: gabelmoo=4060
> https://consensus-health.torproject.org/consensus-health-2016-11-08-18-00.html#36EE8D47E570B8D5515460A9972F3CFD9EDFDFCE: gabelmoo=2110

Yes, measurements are timing and load dependent.

> * authority participating in consensus change, and with only few active (4-5) at a time, impact of a single authority on network is amplified

This is why the median is used.

> By having more servers
> * consensus would be more stable (low-median)
> * leading to more accurate assessment of nodes
> * and wider utilisation of the available nodes
> * leading to higher network throughput
> 
> Does that sound plausible?

The first two, yes, the last two, perhaps.
Maybe lower latency is more important than throughput?

But each bandwidth authority uses bandwidth that would otherwise be available
to clients. So there is a tradeoff.

T

-- 
Tim Wilson-Brown (teor)

teor2345 at gmail dot com
PGP C855 6CED 5D90 A0C5 29F6 4D43 450C BA7F 968F 094B
ricochet:ekmygaiu4rzgsk6n
xmpp: teor at torproject dot org
------------------------------------------------------------------------------



_______________________________________________
tor-relays mailing list
tor-relays@xxxxxxxxxxxxxxxxxxxx
https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays