How to Test a VPS Route to China: Traceroute, MTR, Ping and Peak-Hour Checks

A China-optimized VPS can cost several times more than an ordinary one in the same data center. What you're paying for is the network path between that server and users in mainland China. The trouble is that a plan labeled "CN2 GIA" doesn't guarantee your traffic actually takes CN2 GIA in both directions, and a route that looks fine at 10 a.m. can fall apart at 9 p.m. Beijing time.

This guide shows how to verify the route you bought using free tools, how to read the results by ASN, and how to do it all before the refund window closes. If terms like 163, CN2 GIA, CMI and CMIN2 are new to you, read CN2 GIA VPS for China Access first. Provider terms below are as listed on 2026-10-05.

Why the return path matters

Every connection has two legs:

• Outbound (forward path): your request travels from China to the server.
• Return path: the server sends data back to China.

Requests are small. Pages, files, API responses and video are large, and they all ride the return path. That's why the return path largely determines the speed and stability people in China actually experience.

Routing is often asymmetric. A plan might use ordinary transit outbound and CN2 GIA on the return, or the reverse. So you test both directions, and you weigh the return path more heavily.

The tools

The basic method:

• Outbound: from a computer in China, run mtr <server IP> or tracert <server IP>.
• Return: SSH into the VPS and run nexttrace <your office or home public IP> or mtr <your public IP>.

Use your own public IP as the return-path target. It's the path that actually matters to you, and you're only probing addresses you're responsible for.

Some networks deprioritize ICMP. If results look suspiciously bad, repeat the test in TCP mode; both mtr and NextTrace support it (check each tool's --help for the exact flag on your version).

Reading hops by ASN

The single most useful field in a trace is the ASN (autonomous system number) of each hop. NextTrace shows it inline. With plain traceroute, look each hop IP up in a public BGP lookup such as bgp.he.net.

Telling CN2 GIA from CN2 GT. As a rule of thumb, CN2 hops often sit in 59.43.x.x and 163 backbone hops in 202.97.x.x. If the return path stays on AS4809 the whole way and never touches AS4134, it's generally treated as CN2 GIA. If AS4134 appears partway, you're more likely looking at CN2 GT or an ordinary route. IP prefixes are folklore; the ASN is what counts.

Peak-hour testing and packet loss

Almost every route looks good in the morning. The real test is China's evening peak, roughly 8 p.m. to 11 p.m. Beijing time (UTC+8), when residential traffic surges.

1. Test on three separate evenings, not once.
2. Send enough packets to get a statistic, for example mtr -c 200 <target>, rather than eyeballing a single trace.
3. Keep a daytime run of the same test as a baseline.

How to read mtr loss correctly. If a single intermediate hop shows loss but every hop after it, including the destination, shows none, that router is rate-limiting ICMP replies. It's cosmetic. Real loss starts at one hop and persists all the way to the destination. That first lossy hop is usually where the congestion is.

Watch latency too. A return path whose latency jumps sharply at peak and comes back down later in the night is congested, even if loss stays low.

Before buying: official test IPs and looking glasses

You can do some verification before spending anything:

• Look for a test IP or looking glass. If the provider publishes test IPs, ping and mtr them from China to check the outbound leg. A looking glass lets you run traces from the data center side toward your IP, which approximates the return path. If you can't find either, ask support.
• Read exactly which network the plan uses. For the three VPS providers we list:

AlphaVPS is a Europe/US provider, so there's no China route to verify. Its terms also ban port scanning; stick to ping and traceroute against your own addresses on any provider.

After buying: verify inside the refund window

Run your tests on day one. The windows are short and the conditions vary:

• DMIT: full refund within 3 days with 30 GB or less of traffic used, and a partial refund on new orders within 30 days. Its terms also say there are no refunds for network quality or IP location. Ask support in writing how those two clauses interact before you order.
• BandwagonHost: refund on new orders within 30 days if less than 10% of the transfer allowance has been used.
• AlphaVPS: refund within 14 days on a first order if resources are unused, excluding promo plans, crypto payments and orders from China.

Route tests use almost no bandwidth, so they won't push you over a usage threshold. Just don't migrate production workloads until you've decided to keep the server.

Result cheat sheet

Test only your own servers, your own public IPs and test IPs the provider publishes. For choosing a server in the first place, see Best VPS for Scraping and the VPS comparison table.