Tor is a global anonymity network built on thousands of independently operated relays. Compared with the size of Japan’s Internet infrastructure, however, the country has surprisingly few public Tor relays.
A recent snapshot showed roughly 76 running relays in Japan, with a combined measured relay capacity of about 2.14 Gbit/s. A simple arithmetic calculation gives:
2.14 Gbit/s ÷ 76 ≈ 28.2 Mbit/s per relayThat is only about 3.5 MB/s per relay on average.
This figure needs to be interpreted carefully. Tor relay bandwidth is not evenly distributed. A handful of fast relays can contribute hundreds of megabits per second, while many smaller relays may provide only a few or a few dozen megabits per second. The arithmetic mean therefore does not describe the “typical” relay very well.
A hypothetical distribution might look like this:
A few large relays
300 Mbit/s
200 Mbit/s
150 Mbit/s
Several medium relays
30–80 Mbit/s
Many small relays
5–30 Mbit/sThe faster machines pull the average upward, meaning the median relay could be noticeably slower than the calculated 28.2 Mbit/s.
Relay count is not the whole story
Tor does not route traffic purely according to the number of relays. Directory authorities and bandwidth measurement systems assign weights based on factors such as measured capacity and relay role.
Two countries could therefore both have 100 relays while contributing very different amounts of usable capacity.
A network containing:
100 × 10 Mbit/s relaysis very different from one containing:
100 × 500 Mbit/s relaysThe same distinction applies to consensus weight. A country’s percentage of Tor consensus weight is not simply its percentage of physical relay bandwidth.
Consensus weight reflects the probability that Tor clients will select particular relays for different positions in circuits. It depends on network measurements, relay roles, available bandwidth, and the balancing decisions encoded in each Tor consensus.
Why a small Japanese relay can still matter
From the perspective of the entire Tor network, a relay providing a few tens of megabits per second may appear modest.
From the perspective of Japan, the same relay can be much more significant.
If a country has only several dozen active relays, adding one stable node increases not only capacity but also:
- geographic diversity;
- autonomous-system diversity;
- network-provider diversity;
- operator diversity;
- resilience against failure or concentration.
The diversity aspect can be more important than raw throughput.
For example, adding another relay to a large European hosting provider may increase total bandwidth substantially, but it does little to reduce infrastructure concentration if hundreds of other relays already exist in the same datacenter network.
A smaller independently operated relay in a different Japanese access network can contribute something less easily replicated: another independent routing location.
Residential and non-datacenter relays
A large share of high-capacity Tor infrastructure runs in commercial datacenters and VPS networks. Those environments are attractive because they provide predictable connectivity, static addresses, inexpensive bandwidth, and good uptime.
Residential connections are more complicated. They may involve dynamic addresses, consumer routers, provider policies, and less predictable upload performance.
Nevertheless, a stable residential or small-office relay can add meaningful topology diversity if:
the public address is reachable
+
the connection is stable
+
the operator maintains the system
+
the relay remains online long termA dynamic IPv4 address is not inherently disqualifying. Tor relay identity is primarily associated with its cryptographic identity keys rather than permanently tied to one IP address.
If the public address changes but the relay identity remains intact, the same relay can continue operating after it republishes its descriptor.
Stability usually matters more than short bursts of speed
A relay offering extremely high bandwidth for a few days is less valuable than it may first appear.
A node that stays online continuously for months or years provides predictable capacity and may gradually qualify for additional Tor flags such as Stable, Fast, or eventually Guard, depending on network measurements and authority decisions.
For this reason, a modest but sustainable relay can be more useful than an aggressively configured node that frequently disappears.
A practical philosophy is:
sustainable bandwidth
+
stable uptime
+
consistent identity
+
regular maintenance
>
maximum possible bandwidthFor countries with relatively few relays, this principle becomes even more important.
Japan’s relay population demonstrates an interesting characteristic of Tor: a globally small contribution can still be locally meaningful.
When a country has only a few dozen public relays, every additional stable and independently operated node improves the network in ways that raw global percentages do not fully capture.