Node HandbookNODE HANDBOOK中文
Nodes · region

Choosing a node region: verify the exit, not the label

There is no fixed ranking of regions. A node labelled Tokyo may exit elsewhere, and the region a service actually accepts may not be the one in the name.

Node Handbook editorial team ·

Region choice is not a ranking

For any given task, the region question has two halves: which regions the target service actually accepts, and which node can deliver sustained traffic to that service from your network. Neither half is answered by the provider's naming.

A node name is metadata written by whoever generated the configuration. It can be accurate, stale, or simply aspirational. The checkable facts are the exit IP, the registry data for that IP, and the behaviour of a real request.

Order of judgement for common regions

RegionWhat it is usually chosen forWhat must be verified, not assumed
Hong KongLow latency from mainland and East Asian networksPeak-hour throughput; whether the target service treats it as the region you need
JapanLatency and general web useSustained speed at peak; exit location within the country
SingaporeSoutheast Asian routesActual exit and routing path; congestion patterns
United StatesServices restricted to a US regionWhether the service accepts the specific exit, and whether latency is usable for your task

Treat the right column as the work. The left column is only a reason someone might start there.

Why two people get different results

  • Different local networks have different routes to the same server.
  • Load changes by hour; a result at 03:00 says little about 21:00.
  • Rules differ: one person's request may match a DIRECT rule while the other's is proxied.
  • The target service applies its own policy per account, per IP type, and sometimes per session.

This is why “works for me” and “does not work” can both be true and neither is a general statement about a region.

A region label can be wrong: how to catch it

  1. Connect through the node, then read the exit IP from a service that reports it.
  2. Compare the result against at least two independent IP data sources. They disagree more often than people expect, especially for hosting ranges.
  3. Confirm the target service's own region statement — its help pages, its account region setting, or the behaviour of a region-locked feature.
  4. Record all three: the provider's label, the observed exit, and the service's response.

If the label and the observation disagree, the observation is what you use. If the observation is ambiguous, write “unverified” rather than picking the answer you wanted.

Requirement first, distance second

Settle which regions the service supports before comparing latency at all. A 40 ms node in an unsupported region loses to a 180 ms node in a supported one, for that task. Distance is a hint about round-trip time; it does not predict whether an HTTPS request completes or whether a stream sustains bitrate.

What to record when comparing regions

  • Time, network type and client, so a later comparison is possible.
  • Node label, protocol, and observed exit IP.
  • Result of a real HTTPS request to the actual target, not a generic page.
  • Latency plus at least one throughput observation, ideally at peak hour.
  • The service's response: accepted, region error, or account restriction.

One comparison at one moment is one data point. Do not promote it to “this region is stable”.

Scope

This page describes a verification order, not a provider ranking. It contains no speed testing of specific services and makes no claim about which region is best in general — the answer depends on the service you need and the network you are on.