Wow, thank you all for the suggestions!
Hope to implement these soon. Would definitely appreciate more ideas too.
Sean
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA512
Wow, thank you all for the suggestions!
Hope to implement these soon. Would definitely appreciate more ideas too.
If you have the time to maintain such a factor you could weight in relative bw cost per region, because the prices per bw unit heavily deffer depending on the region in which a relay is running.
e.g. relay runs in South America -> expensive bw -> multiplier greater than 1
I'm not sure if there any public lists that you could regularly fetch to automate this.
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On Thu, 25 Jun 2015 10:49:30 +0200 nusenu nusenu@openmailbox.org wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA512
Wow, thank you all for the suggestions!
Hope to implement these soon. Would definitely appreciate more ideas too.
If you have the time to maintain such a factor you could weight in relative bw cost per region, because the prices per bw unit heavily deffer depending on the region in which a relay is running.
e.g. relay runs in South America -> expensive bw -> multiplier greater than 1
A relay running in South America could do more bad than good, as it would increase the average latency if e.g. users in Europe accessing sites in Europe now have to go via South America on some circuits.
However most likely it won't get a high enough consensus weight in the first place, e.g. I run a relay in Japan on a gigabit connection, but nobody cares too much, since (I assume) bwauths aren't anywhere near Japan and do not get good speeds to it, they give it a low weight, and as a result it doesn't see a lot of use.
- -- With respect, Roman
A relay running in South America could do more bad than good, as it would increase the average latency
I was also thinking about that. "Does improving geo-diversity negatively affect latency?"
and I agree with you that the example (more relays in South America) would probably increase avg. latency for European users but at the same time has a change to decrease latency for e.g. Brazilian users.
So in the end the question is where are most of your users? And what is your optimization goal/what is your priority? (a low worst-case latency for all users or a low latency for most of the users) https://metrics.torproject.org/userstats-relay-table.html
I would still prefer having more relays in uncommon locations even if that comes with an (acceptable) latency penalty. Whatever 'acceptable' is.
Since the question "How does a good relay look like and where is it located?" is a non-trivial one and because Roaster is effectively codifying the (Torproject's?) answer to it I'm also interested in the outcome from this point of view.
On Thu, Jun 25, 2015 at 6:27 AM, nusenu nusenu@openmailbox.org wrote:
A relay running in South America could do more bad than good, as it would increase the average latency
I was also thinking about that. "Does improving geo-diversity negatively affect latency?"
Internet imposed minimum latency to exit destinations in geophysical worst case antipodal node distribution is roughly 1000ms, or 500ms with average distribution. HS destinations are about double that.
Putting nodes in nontraditional locations could actually lower average latency because they could enable a more direct fiber route / waypoint than BGP arcing by default back through some longer path along the densely laid areas.
Tests could be done with manual circuits and global fiber maps.
Todays node distribution is probably already subject to the random path minimum averages above, so I doubt it would hurt. Plus you get more geopolitical diversity.
Play pin the tail on the donkey and win a prize for being the first in an unoccupied country / state.
tor-relays@lists.torproject.org
-
grarpamp -
nusenu -
Roman Mamedov -
saitosean＠ymail.com