Full-Throttle Tor: Real-World 10 Gbps Performance
If you're a software developer who cares about freedom of information—or an IT ops admin with a spare rack and skills to spare—this post is for you. Let's talk about how you can use your talents to support the Tor network.
Summary Highlights
- New York and Chicago exhibited the highest average bandwidth per relay.
- San Jose consistently had the lowest average bandwidth per relay.
- Detroit and Dallas showed slow but steady increases, though they lag behind New York and Chicago.
- Overall bandwidth utilization averaged around 50% of the purchased capacity (~16 Gbps of 30 Gbps).
- Most relays have been active for less than two months; longer 'brew' times (3–6 months) often yield more stable bandwidth patterns.
Next Steps
- Discontinue underperforming relays in Detroit and Dallas (Server 2).
- Begin deployments with a fifth provider in New Jersey (Secaucus) and Montreal.
- Consider adding extra relays in New York and Chicago based on the new provider's performance.
Detailed Analysis
New York saw the fastest growth, rising from 7.38 Mbps to 8.70 Mbps over three weeks. Chicago improved steadily, while Detroit’s performance increased modestly. Dallas experienced mixed results, and San Jose’s lower performance is likely related to further distance from US East / Europe where there's more Tor traffic allocated.
Temporary dips in bandwidth corresponded to migration from rented IPv4 ranges to our own prefixes. All servers run Ubuntu 24.04.02 with the latest Tor versions, using 88–192 CPU threads and 384–1000 GB RAM on unmetered 10 Gbps connections. Each server costs approximately $500/month.
Average Bandwidth per Relay (Mbps)
| Provider (Rank) | Apr 26th 2025 | Apr 19th 2025 | Apr 12, 2025 | Apr 3, 2025 | Mar 30, 2025 | Mar 22, 2025 |
|---|---|---|---|---|---|---|
| 1 – New York | 6.4 | 7.8 | 8.70 | 8.25 | 7.91 | 7.38 |
| 2 – Chicago | 7.6 | 7.3 | 8.22 | 5.44 | 4.73 | 4.24 |
| 5 – New Jersey | 6.8 | 5.6 | 6.1 | 0 | 0 | 0 |
| 5 – Montreal | 5.0 | 5.4 | 4.2 | 0 | 0 | 0 |
| 1 – Detroit | 0 | 0 | 5.25 | 5.17 | 4.55 | 4.60 |
| 3 – Dallas 1 | 4.8 | 4.7 | 4.95 | 2.82 | 2.46 | 4.34 |
| 3 – Dallas 2 | 0 | 0 | 4.73 | 2.84 | 2.76 | 3.52 |
| 4 – San Jose | 4.2 | 3.5 | 3.38 | 3.90 | 3.26 | 3.06 |
Chart: Average Bandwidth per Relay Over Time
Community Feedback & Recommendations
Another Tor relay contributor, John Crow highlighted that New York and Chicago are major hubs (often Secaucus, NJ for “New York”). He recommended exploring Ashburn, VA—an established Internet hub with robust US and EU connectivity—and Miami, FL, which offers strong connectivity to South America. He also noted San Jose’s underperformance likely stems from additional hops through Los Angeles.
Conclusion
Early data suggests that while New York and Chicago deliver the best bandwidth per relay, alternative locations like Ashburn and Miami warrant testing. Patience is crucial; as relays “brew” beyond two months, performance patterns often stabilize. We’ll continue refining our deployments and welcome further community insights.
For full details and discussion, see the Tor Project mailing list thread.
Join the Mission