Hello,
New York and Chicago are both major hubs so it makes sense why they have highest average bandwidth. Keep in mind, When providers say New York most of the time it is just Secaucus, NJ. There are multiple datacenter right next to each other.
I would recommend looking at some other locations you should look at is Ashburn, VA and Miami, FL. Ashburn is already a well established hub in the US due to the growing number of cloud datacenter located there. It has excellent connectivity to both the rest of US and EU. In addition, it also helps alleviate the congestion between NY/NJ and EU. Miami has lately been one of the up and coming locations because it is connected to US, South America and EU via NY. Both of these locations seem to do well for Tor.
With San Jose, it makes sense why it is underperforming. Often times traffic has an extra hop because it all has to go through Los Angeles.
On Wed, Apr 16, 2025, at 12:03 AM, Tor at 1AEO via tor-relays wrote:
Summary: New York and Chicago locations had the highest average bandwidth per relay. San Jose had the consistently lowest average bandwidth per relay. Dallas and Detroit are slowly increasing, but hard to see the same potential of New York and Chicago. Main constraint preventing broader evaluation is the ~360 relays per family: https://gitlab.torproject.org/tpo/core/tor/-/issues/40837
Next steps: Discontinuing Detroit and Dallas - Sever 2. Starting servers with a 5th provider in New Jersey and Montreal. Tempted to add an extra server in New York and Chicago, but will see how 5th provider goes first.
Details: New York increased the fastest and has stayed at number one. Chicago had steady increases but recently had significant increase. Detroit has slow and steady increase. If Chicago wasn't doing so much better, would have Detroit operate longer. Dallas seems a very slow increase. The dips were due to two moves from rented IPv4 to owned IPv4. Will give it another month or so and see where it ends up. Disappointed in San Jose consistently at the bottom.
Overview of data and servers: Covering first ~2 months of ~340 guard relays across 4 different providers in 5 different US-based geographic locations. All relays with roughly the same average life of ~2 months by the April date. The servers are all 88 threads to 128 threads with 384GB to 768GB RAM and each on 10 Gbps dedicated unmetered connections. The servers are all running Ubuntu 24.04.02 with latest Tor relay versions. Half the servers using owned IPv4 with my ASN and half using rented IPv4 with provider ASN. Each server is roughly $500/mo and majority are privately negotiated because list prices are 2-5x higher. Every provider said they're comfortable with Tor traffic but most weren't familiar with the details. Source: All data from metrics.torproject.org
Table and chart attached as screenshots and text table pasted below.
Table in text: Average Bandwidth per Relay Provider Apr 12th 2025 Apr 3rd 2025 March 30th 2025 March 22nd 2025 1 - New York 8.70 8.25 7.91 7.38 2 - Chicago 8.22 5.44 4.73 4.24 1 - Detroit 5.25 5.17 4.55 4.60 3 - Dallas - Server 1 4.95 2.82 2.46 4.34 3 - Dallas - Sever 2 4.73 2.84 2.76 3.52 4 - San Jose 3.38 3.90 3.26 3.06
For Felix, most relays have HsDir flag and average uptime for a week or two or three, depending on when last Tor version bumped. Likely when you review relays had just been upgraded so HsDir flag often drops for a few days.
On Saturday, April 5th, 2025 at 1:03 PM, Felix via tor-relays tor-relays@lists.torproject.org wrote:
Hi
Part 3 - Looking for feedback on our Tor relay configuration decision:
Looking for guidance on optimizing server / bandwidth spending for Tor guard / middle relays across datacenters.
2Gbps means 1Gbps in and 1Gbps out.
Thanks for the informations.
Overall, using 16 Gbps out of 30 Gbps purchased -- roughly half our bandwidth is idle / "wasted"
Is it right that most (all?) relays are not older than mid of March? If so please take a look at https://blog.torproject.org/lifecycle-of-a-new-relay/ .
As for as I see the relay status page shows not a single relay with a HsDir flag and no uptime above 2 days.
Gaining a trustful relay with good traffic needs much more time alive of the relays (keys). Easily 3 to 6 months.
The relays are on up to date Tor versions 4.8.15 and 4.8.16 what is good!
My recommendation is to let them brew for 2 months. The situation will change.
Cheers and happy relaying, Felix
tor-relays mailing list -- tor-relays@lists.torproject.org To unsubscribe send an email to tor-relays-leave@lists.torproject.org
tor-relays mailing list -- tor-relays@lists.torproject.org To unsubscribe send an email to tor-relays-leave@lists.torproject.org
*Attachments:* • image.png • image.png