China Traffic Routing 101 (2026)

China traffic routing explained (2026): why China latency is different, how Telecom/Unicom/Mobile routes work, what 'good BGP' really means, and how to choose regions, carriers, and CDNs with confidence.

China Traffic Routing 101 (2026)

If your users are in Mainland China, latency and stability are not determined by geography.

They’re determined by:

  • Which Chinese carrier your users are on (Telecom / Unicom / Mobile)
  • What international routes your server’s upstream uses (BGP quality)
  • Whether peak-hour congestion hits your chosen path
  • Whether your traffic is static (CDN-friendly) or dynamic (routing-sensitive)

This guide gives you a mental model and a repeatable decision process.

1️⃣ Executive Summary (The “Truth”)

If you remember only five rules:

  1. Hong Kong is not automatically fast — cheap HK can be worse than Tokyo.
  2. BGP quality beats distance — “CN2/CMI-grade” routes are what you’re buying.
  3. Static vs dynamic matters — CDN helps static; routing quality matters for dynamic.
  4. Test on three carriers — Telecom + Unicom + Mobile behave differently.
  5. Peak hour decides reality — test during China evening congestion (not daytime).

2️⃣ Why China Traffic Behaves Differently

China’s outbound traffic typically exits through controlled international gateways and major carrier backbones.

Practical consequences:

  • Some routes are stable but expensive
  • Some routes are cheap but congested
  • Some routes look fine off-peak and collapse at night

If you serve China traffic, you’re not just buying “a server.”
You’re buying a routing relationship.

3️⃣ The Three Major Carriers (CT / CU / CM)

China Telecom (CT)

  • Large footprint, strong in many regions
  • Often the “baseline” for enterprise traffic
  • Quality varies heavily by upstream and peering

China Unicom (CU)

  • Can be excellent in some north/metro routes
  • Often more predictable in certain paths than cheap CT routes

China Mobile (CM)

  • Huge user base
  • International routing can be “different,” sometimes unstable on cheaper upstreams
  • If you ignore Mobile, you can lose a large segment of real users

Important:
A region that is great for Telecom may be mediocre for Mobile.
That’s why “single test” benchmarks lie.

4️⃣ What “Good Routing” Actually Means (BGP Quality Tiers)

You’ll see marketing terms like:

  • CN2
  • CMI
  • Premium
  • Optimized
  • China route

They’re not standardized. You need a model.

Tier Model (Decision-first)

Tier A — Premium China-friendly routing

  • Designed to avoid bad congestion paths
  • More stable peak-hour performance
  • Costs more

Tier B — Decent mainstream transit

  • Works fine for some users
  • Peak hour may degrade
  • Often acceptable if your app is not latency-critical

Tier C — Cheap transit (the trap)

  • Fine in tests at 2pm
  • Falls apart at 9pm
  • Packet loss and jitter show up unpredictably

You want Tier A or good Tier B depending on your scenario.

5️⃣ Region Choice (Reality, not map distance)

This is the region model most teams eventually converge on:

🇭🇰 Hong Kong

Best when:

  • Your upstream is premium and proven for China routes
  • You care about the lowest possible RTT

Risk:

  • Cheap HK is a frequent disappointment

🇯🇵 Tokyo

Best when:

  • You want predictable stability for broad China audiences
  • You want a “safe default” that often outperforms cheap HK in practice

🇸🇬 Singapore

Best when:

  • Budget is important
  • You target South China / SEA mixed traffic
  • You accept higher RTT for North China

🇺🇸 US West

Best when:

  • You need global architecture and China is one segment
  • You use CDN heavily and dynamic traffic is limited
  • You need backup/failover

6️⃣ Static vs Dynamic: Where CDN Helps (and where it doesn’t)

CDN is great for:

  • images, CSS/JS, downloads
  • static landing pages
  • caching-heavy e-commerce assets

CDN does NOT solve:

  • login flows
  • checkout
  • API responses that can’t be cached
  • realtime apps (voice, gaming)
  • WebSocket-heavy products

If your site is dynamic-heavy, routing quality matters more than CDN.

7️⃣ Pricing Reality Breakdown (What You Actually Pay For)

Your “China performance cost” typically comes from:

  • premium routing / better transit upstreams
  • bandwidth (especially egress)
  • CDN usage (if you go heavy static)
  • multi-region duplication (if you build failover)

Cheap server + bad routing = expensive in lost conversions.

8️⃣ Deployment Patterns That Actually Work

Pattern 1: “Tokyo + CDN”

Best for:

  • WordPress
  • content-heavy sites
  • SaaS with mixed traffic

Why it works:

  • stable baseline routing
  • CDN absorbs static traffic

Pattern 2: “Hong Kong premium + CDN”

Best for:

  • latency-sensitive China-first products
  • e-commerce with frequent interactions
  • businesses where China conversion matters

Why it works:

  • lowest RTT + stable routing when premium is real

Pattern 3: “HK + Tokyo failover”

Best for:

  • mission-critical products
  • higher-budget teams

Why it works:

  • hedge against route degradation
  • allows carrier-specific optimization

Pattern 4: “Mainland China hosting (ICP)”

Best for:

  • strict compliance needs
  • maximum China performance
  • deep China business operations

Cost:

  • compliance + ICP work + operational overhead

9️⃣ How to Test Routing Properly (Step-by-step)

What you measure

  • RTT (latency)
  • jitter (stability)
  • packet loss
  • traceroute path (where congestion happens)
  1. Deploy a test VM in target region
  2. Run traceroute from China Telecom + Unicom + Mobile sources
  3. Run tests at:
    • off-peak (daytime)
    • peak (China evening)
  4. Repeat for 2–3 days
  5. Decide based on consistency, not best-case results

Key insight:
Peak hour performance is the true product you’re buying.

10️⃣ Common Mistakes (Costly ones)

❌ Choosing by map distance
❌ Believing “CN2” claims without traceroute validation
❌ Only testing from one ISP
❌ Testing once and deploying at scale
❌ Overusing CDN and ignoring dynamic latency
❌ Ignoring packet loss (loss kills UX more than RTT)

11️⃣ FAQ (12)

1) Is Hong Kong always the best for China?

No. Only if routing quality is premium and proven.

2) Is Tokyo a safer default?

Yes. Tokyo often wins on consistency.

3) Do I need CN2 for every project?

No. Only when you need consistent peak-hour performance.

4) Is Singapore OK for China?

Often yes for budget setups, but RTT to North China is higher.

5) Can CDN solve China performance?

Only for static-heavy sites. Dynamic performance still depends on routing.

6) Why do daytime tests look great but nights are terrible?

Peak-hour international congestion.

7) Should I use Mainland China hosting?

Only if you accept compliance and operational overhead (ICP etc.).

8) What matters more: latency or packet loss?

Packet loss. Loss destroys UX even at low RTT.

9) Is “premium route” a standard label?

No. Always validate via traceroute and multi-carrier tests.

10) What’s the quickest way to decide region?

Tokyo + CDN is the least-regret baseline for many cases.

11) What’s the most accurate testing method?

Multi-carrier tests at peak hour across multiple days.

12) How do I avoid wrong purchases?

Treat routing as a product. Test before committing.

12️⃣ Next Steps (Internal Links / Closed Loop)