The three options side by side
| Overseas origin, direct | Near-shore edge | Mainland hosting, filed | |
|---|---|---|---|
| Filing required | No | No | Yes |
| Time to launch | None | Hours | Weeks |
| Typical round trip from major mainland cities | 150-300 ms | 20-60 ms to the edge | 10-40 ms |
| Evening peak behaviour | Degrades, with loss and handshake failures | Mostly unaffected, because cache hits never cross the border | Unaffected |
| Where the data sits | Your existing origin region | Your existing origin region | Inside the mainland |
| Certificates | Yours to manage | Typically issued and renewed by the edge | Yours, with local constraints |
| Main failure mode | The cross-border path | Origin reachability from the edge | Provider and filing maintenance |
What each option is good at
Overseas origin, direct. Cheapest and simplest, and entirely adequate when China is a rounding error in your traffic. If nobody has complained, this is fine, and it is a reasonable place to start.
Near-shore edge. The best ratio of improvement to effort. Most requests never leave the edge, so the cross-border path stops being on the critical path for the common case, and it works with the origin you already run.
Mainland hosting. The best latency available, and the only option when data has to stay in the country. It also carries the most work that is not technical: the filing, ongoing compliance, and a provider relationship you cannot avoid.
Where the trade-off actually bites
- How much of your site is cacheable. The more static the site, the better the edge performs and the less it matters where the origin is.
- How much traffic is genuinely dynamic. An API returning unique data on every call gets far less from an edge than a marketing site does.
- Whether you have long-lived connections. Sessions and streams are the most exposed to cross-border instability, and where a near-shore edge earns the most.
- Your regulatory posture. If your sector carries data residency obligations, the decision is already made and only one column is available.
- Organisational cost. A mainland filing is a recurring administrative commitment, not a one-off project.
The switching-cost question
Ask this before choosing: how much work is it to change our mind in a year?
If the hostname stays the same and the DNS record is one CNAME, changing your mind is a DNS change. If moving means a new domain, new certificates, and a redirect from the old one, then the first choice you make becomes permanent, which is a good argument for keeping the hostname independent of the infrastructure underneath it.
How to decide
- China traffic is small and nobody is complaining: change nothing.
- China users are a real segment and you have no mainland entity: near-shore edge.
- China users are the primary segment, or data must stay in-country: mainland hosting with a filing.
- You are not sure yet: near-shore edge now, with the migration path kept open.