Speed Test Nodes: 32 Nodes, 10 Datacenters, 4 Line Classes — and How to Use Them
The coverage map of all test nodes, the four line tiers inside one datacenter, browser and command-line measurement recipes, and the honest scope of each numb
# What this list is
The test-node inventory is the ammunition depot for evidence-based selection: **32 test nodes across 10 datacenters and 4 line classes**, one click away in the browser, and copy-paste IPs for the command-line crowd who want a full ping/mtr/iperf3 evidence chain. The live, complete node and IP list lives on the online speed test page (click any IP to copy); this guide explains the structure and how to use it honestly.
## 10 datacenters: the coverage map
| Region | Code | Main service direction |
|---|---|---|
| Hong Kong | HKG | Nearest point for South-China users; international + CN2 samples |
| Singapore | SIN | SEA hub; southbound users and transshipment business |
| Tokyo | NRT | Balanced point for East/North China; rich international exits |
| Seoul | ICN | NE-Asia comparison point |
| Kuala Lumpur | KUL | SEA second-tier comparison |
| Taipei | TPE | East-Asia near-end comparison |
| Frankfurt | FRA | Europe business and CE interconnect hub |
| Los Angeles | LAX | US-West main口岸 for China direction |
| Seattle | SEA | US-West comparison point |
| Silicon Valley | SJC | US-West tech corridor, dense CN–US cable landings |
One selection rule: **test the direction your real users are on.** If your users are Shanghai Telecom, measure LAX/HKG on the Telecom direction — not your own datacenter's local LAN. Method and acceptance criteria are in the speed-test methodology article.
## 4 line classes: quality tiers inside one datacenter
- **International BGP** — standard multi-carrier peering; the price baseline (our 1G tier at $184.80/mo).
- **Enterprise** — guaranteed tier aimed at business customers.
- **CN2 China-optimized** — Telecom's premium international channel direction (our 1G tier at $433.20/mo); evening-peak stability clearly better than standard channels. The premium buys exactly that.
- **China-optimized VIP** — higher-grade CN direction optimization (CN2 GIA-class resources, $1047.60/mo tier) for experience-first businesses.
Same datacenter, different line, same evening peak: the delta is direct evidence of whether a premium is worth paying. Line theory and selection criteria live in the routing hub.
## Browser method: three steps
1. **Auto-detect.** Open the speed test tool; it identifies your carrier (Telecom/Unicom/Mobile/edu/overseas) and nearest access point, then orders nodes by recommendation. Wrong guess? Override manually.
2. **Pick a mode.** Full test ≈ 10 MB (latency + jitter + probe failure rate + download); latency-only ≈ 50 KB (recommended on mobile).
3. **Read it right.** Latency = median; stability = jitter; availability = probe failure rate; download speed is capped by *your* last mile — use it for node-vs-node comparison only. Exact definitions live in the methodology section.
## Command-line method: full evidence chain
```bash
TESTIP=
# 1) latency distribution (20 packets: median + tail)
ping -c 20 $TESTIP | tail -2
# 2) per-hop path and loss (the bidirectional core, 100 packets)
mtr -rwzbc 100 $TESTIP > mtr-$(date +%F-%H%M).txt
# ICMP rate-limited? switch to TCP/443
mtr -T -P 443 -rwzbc 100 $TESTIP > mtr-tcp-$(date +%F-%H%M).txt
# 3) throughput via the public test file
curl -o /dev/null -sw "speed: %{speed_download} B/s\n"
# 4) HTTP phases (DNS / connect / TTFB / total)
curl -o /dev/null -sw "dns:%{time_namelookup} conn:%{time_connect} ttfb:%{time_starttransfer} total:%{time_total}\n" https://axionserve.com/
# 5) repeat 1, 2, 4 at evening peak (20:00-23:00) and archive beside the off-peak run
```
## Using node data in selection
- **Pre-purchase verification:** target datacenter × your users' direction × evening peak — three measured rounds before ordering. "Test first, buy second" is the entire reason this tool exists.
- **Tier comparison:** measure International BGP and the CN-optimized tier in the same datacenter at the same peak; then answer with your own numbers whether the $184.80 → $433.20 gap is worth the loss/latency improvement.
- **Monthly archive:** one same-slot round per month gives you a personal time series of line quality — which is exactly why the monthly line digest welcomes reader data.
- **Honest scope:** browser probes currently measure via the Cloudflare edge (disclosed on the tool page), i.e. "you → edge" quality; node-direct domains switch to end-to-end measurement when live. The two scopes are never mixed in one statistic.
## FAQ
**Why is my latency so different from a review article?** Carrier, time slot, path — change any one and numbers stop being comparable. Compare against your own three rounds, not someone's screenshot.
**Do node IPs change?** The tool page is the source of truth; this article deliberately does not mirror the IP list, so it can never go stale.
**Why doesn't download hit the advertised bandwidth?** Browser tests are capped by your last mile, single-stream TCP and Wi-Fi. For true port throughput use multi-thread iperf3 (templates in the MTR & Looking Glass article).
Related tools: measurement methodology · MTR & Looking Glass · bandwidth-and-cost calculator. Entry point: online speed test; the diagnostics panorama lives in the speed-test & diagnostics hub; datacenter choice behind the nodes lives in the live configurations page.