I'm standing up a new exit relay on the VULTR network. How would a person go about determining which location is in most need of additional exit relay capacity?
Available locations: https://www.vultr.com/locations/
* Miami, Florida * Chicago, Illinois * New York / New Jersey * Dallas, Texas * Seattle, Washington * Atlanta, Georgia * Los Angeles, California * Silicon Valley, California * (AU) Sydney, Australia * (Asia) Tokyo, Japan * (EU) Amsterdam, NL * (EU) London, UK * (EU) Paris, France * (EU) Frankfurt, DE
Also, curious to hear people's thoughts on any potential jurisdictional arbitrage benefits to be gleaned by choosing a location other than ones country of residence or citizenship.
For the sake of argument, consider a VULTR account opened by U.S. citizen residing in the U.S. Choopa LLC (VULTR parent company) is also a US based company. http://start.cortera.com/company/research/k5o8lvm2j/choopa-llc/
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA512
Hi Seth,
I'm standing up a new exit relay on the VULTR network. How would a person go about determining which location is in most need of additional exit relay capacity?
thanks for taking network diversity into account when setting up new relays!
It might be oversimplified but using compass with group by country ordered by consensus weight (or in your case exit probability) shows you where most of tor network capacity is currently located. The goal is to setup relays in new or rarely used locations.
So by using compass your list would look like this, ordered from better to less good:
* (AU) Sydney, Australia (0.01% CW) * (Asia) Tokyo, Japan (0.8% CW) * UK (4.6% CW) * US (10.1%) * NL (12.4% CW) * France (21.6%) * DE (25.7% CW) Note: the is a current snapshot and numbers change but AU or JP is better then DE (from a capacity divers. point of view) - this will also be the case in a week or a month.
You might also want to consider the exit probability and use that in addition or instead of CW.
I don't know if VULTR has multiple ASes but if they do you might also want to have a look at the group by AS results (if they allow you to choose).
On Fri, 01 May 2015 10:01:45 -0700, nusenu nusenu@openmailbox.org wrote:
It might be oversimplified but using compass with group by country ordered by consensus weight (or in your case exit probability) shows you where most of tor network capacity is currently located. The goal is to setup relays in new or rarely used locations.
So by using compass your list would look like this, ordered from better to less good:
- (AU) Sydney, Australia (0.01% CW)
- (Asia) Tokyo, Japan (0.8% CW)
- UK (4.6% CW)
- US (10.1%)
- NL (12.4% CW)
- France (21.6%)
- DE (25.7% CW)
Note: the is a current snapshot and numbers change but AU or JP is better then DE (from a capacity divers. point of view) - this will also be the case in a week or a month.
You might also want to consider the exit probability and use that in addition or instead of CW.
I don't know if VULTR has multiple ASes but if they do you might also want to have a look at the group by AS results (if they allow you to choose).
Thanks for the breakdown, that helps. The only hitch with the Sydney and Toyko locations is that instead of 1000GB/mo of bandwidth, you only get 200GB/mo.
Would it be better (all things considered) to go with the UK location at 1000GB/mo vs Tokyo or Sydney at 200GB/mo?
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256
I would love to see some more nodes in Australia. I'm located in Perth and the speed of the network it horrible. Not usable for day to day internet which is unfortunate, hopefully it will pick up soon.
I might look into setting up a node here as my only running one is located out of Australia
On 2 May 2015 at 05:00, Seth list@sysfu.com wrote:
On Fri, 01 May 2015 10:01:45 -0700, nusenu nusenu@openmailbox.org wrote:
It might be oversimplified but using compass with group by country ordered by consensus weight (or in your case exit probability) shows you where most of tor network capacity is currently located. The goal is to setup relays in new or rarely used locations.
So by using compass your list would look like this, ordered from better to less good:
- (AU) Sydney, Australia (0.01% CW)
- (Asia) Tokyo, Japan (0.8% CW)
- UK (4.6% CW)
- US (10.1%)
- NL (12.4% CW)
- France (21.6%)
- DE (25.7% CW)
Note: the is a current snapshot and numbers change but AU or JP is better then DE (from a capacity divers. point of view) - this will also be the case in a week or a month.
You might also want to consider the exit probability and use that in addition or instead of CW.
I don't know if VULTR has multiple ASes but if they do you might also want to have a look at the group by AS results (if they allow you to choose).
Thanks for the breakdown, that helps. The only hitch with the Sydney and Toyko locations is that instead of 1000GB/mo of bandwidth, you only get 200GB/mo.
Would it be better (all things considered) to go with the UK location at 1000GB/mo vs Tokyo or Sydney at 200GB/mo?
tor-relays mailing list tor-relays@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays
On Sat, 02 May 2015 00:52:07 -0700, Geo Rift tim.cochrane.laptop@gmail.com wrote:
I would love to see some more nodes in Australia. I'm located in Perth and the speed of the network it horrible.
Tim, just deployed an exit node to Sydney location, feel free to test it out:
https://atlas.torproject.org/#details/E1E1059D8C41FC48B823C6F09348EA89C4D4C9...
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256
Great to see Seth! I'm looking around Perth area at the moment for cheaper bandwidth but can't seem to find anything near that kind of price. A bit of a shame but it's the reality given Australia's poor internet.
On 4 May 2015 at 12:35, Seth list@sysfu.com wrote:
On Sat, 02 May 2015 00:52:07 -0700, Geo Rift < tim.cochrane.laptop@gmail.com> wrote:
I would love to see some more nodes in Australia. I'm located in Perth and the speed of the network it horrible.
Tim, just deployed an exit node to Sydney location, feel free to test it out:
https://atlas.torproject.org/#details/E1E1059D8C41FC48B823C6F09348EA89C4D4C9...
tor-relays mailing list tor-relays@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA512
You might also want to consider the exit probability and use that in addition or instead of CW.
I don't know if VULTR has multiple ASes but if they do you might also want to have a look at the group by AS results (if they allow you to choose).
Thanks for the breakdown, that helps. The only hitch with the Sydney and Toyko locations is that instead of 1000GB/mo of bandwidth, you only get 200GB/mo.
Would it be better (all things considered) to go with the UK location at 1000GB/mo vs Tokyo or Sydney at 200GB/mo?
Is there a specific reason why you limit yourself to vultr?
On Sat, 02 May 2015 14:37:04 -0700, nusenu nusenu@openmailbox.org wrote:
Is there a specific reason why you limit yourself to vultr?
Yes, there are several.
* Price (hardware bang for the buck. SSD, 1000GB bw/mo in most locations. Starter pkg is $5/mo) * Features/usability (really like their control panel and website design. Snapshots are key, ability to re-deploy snapshots anywhere. Two factor auth with Yubikey.) * OpenBSD supported via custom ISO install feature (This limits the field quickly)
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA512
- Price (hardware bang for the buck. SSD, 1000GB bw/mo in most
locations. Starter pkg is $5/mo)
I'd say 7$ for 2TB/mo on 1GB RAM is expensive if you compare it with 100mbps unmetered and lets say you are able to saturate ~50% = ~30TB/mo (~50 mpbs* in one direction) for ~15$/mo with 1GB RAM (in HU, 0.6% CW).
https://lists.torproject.org/pipermail/tor-relays/2013-January/001835.html
2TB/mo is not a lot of traffic, it translates to less than 4mpbs in one direction.
..but anyway thanks for adding more OpenBSD relays.
OVH is pretty good value,
CAD$2.99/mo for 1GB RAM and unlimited transfer at 100Mbps (it’s speed limited after 10,000GB) and both IPv4/6.
However there are 424 OVH relays across 12 countries might not fit with your goal to add more diversity https://compass.torproject.org/#?exit_filter=all_relays&links&sort=c... https://compass.torproject.org/#?exit_filter=all_relays&links&sort=cw&sort_reverse&country=&ases=AS16276&by_country&top=-1&by_as=false
On May 3, 2015, at 2:50 PM, nusenu nusenu@openmailbox.org wrote:
Signed PGP part
- Price (hardware bang for the buck. SSD, 1000GB bw/mo in most
locations. Starter pkg is $5/mo)
I'd say 7$ for 2TB/mo on 1GB RAM is expensive if you compare it with 100mbps unmetered and lets say you are able to saturate ~50% = ~30TB/mo (~50 mpbs* in one direction) for ~15$/mo with 1GB RAM (in HU, 0.6% CW).
https://lists.torproject.org/pipermail/tor-relays/2013-January/001835.html
2TB/mo is not a lot of traffic, it translates to less than 4mpbs in one direction.
..but anyway thanks for adding more OpenBSD relays.
tor-relays mailing list tor-relays@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA512
OVH is pretty good value,
CAD$2.99/mo for 1GB RAM and unlimited transfer at 100Mbps (it’s speed limited after 10,000GB) and both IPv4/6.
However there are 424 OVH relays across 12 countries might not fit with your goal to add more diversity
Yes, OVH AS is probably the worst place to add relays from a diversity pov since it is the AS with the highest CW fraction (>10%).
On Sun, 03 May 2015 11:50:25 -0700, nusenu nusenu@openmailbox.org wrote:
I'd say 7$ for 2TB/mo on 1GB RAM is expensive if you compare it with 100mbps unmetered and lets say you are able to saturate ~50% = ~30TB/mo (~50 mpbs* in one direction) for ~15$/mo with 1GB RAM (in HU, 0.6% CW).
Can't argue with that.
The difference in annual cost ($60 vs $180 USD) is the key factor for me right now. Don't want to pay $180/yr out of pocket right now.
..but anyway thanks for adding more OpenBSD relays.
Aye, I'll be trying out your Ansible playbooks in a bit.
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA512
..but anyway thanks for adding more OpenBSD relays.
Aye, I'll be trying out your Ansible playbooks in a bit.
Glad to hear. Let me know if you run into any issues. The way openfiles-max is set should definitely get some testing.
Note that I changed the way tor is installed on OpenBSD today, since packages are outdated (0.2.5.10) ansible-relayor installs from ports now - which requires ports to be available on the system (see dependencies), but I'm not planing to keep that dependency for long (the role should take care of that as well but it will be opt-in since I don't want to mess with people's ports tree out of the box).
See also: https://github.com/nusenu/ansible-relayor/issues/26
tor-relays@lists.torproject.org
-
Ethan Rose -
Geo Rift -
nusenu -
Seth