Round-trip time from AWS eu-south-2 (Spain / Aragón) to every other AWS commercial region.
| # | City | Region | IATA | Latency ▼ | Band |
|---|
Want the whole picture? See the full 35 × 35 region-to-region matrix →
https://www.cloudping.co/api/latencies?percentile={p_10|p_50|p_99}&timeframe=1Wp10 is the fast-path (10% of samples are faster). p50 is the median. p99 is the tail — 1 in 100 samples is at least this slow. Design SLAs against p99, not p50.us-gov-*) and China (cn-*) regions are isolated environments and cannot be probed from commercial accounts, so they are absent by design.me-south-1) shows identical p10/p50/p99 — a sparse-sampling artifact in the source, not a suspiciously stable link.curl 'https://www.cloudping.co/api/latencies?percentile=p_50&timeframe=1W' | jq '.data["eu-south-2"]'Six public sources were evaluated. Only one is used here — the others measure a different thing, have no API, or are dead.
| Source | Method | API | Verdict |
|---|---|---|---|
| CloudPing.co | EC2 → EC2 TCP, all 35 regions | Yes, JSON | Used here. Only public source with p10/p50/p99, 1D/1W/1M/1Y windows, and full N×N coverage. |
| awsspeedtest.com | Browser → S3 HTTPS | No | Client-to-region only. Adds ~20–40 ms of browser/TLS overhead vs. EC2 probes. Good for a laptop-side sanity check. |
| cloudping.info | Browser HTTP ping | No | Single-shot only. Not the Adorjan tool — that's .co. Fine for one-off checks, no history. |
| cloudpingtest.com | Browser HTTPS, multi-cloud | No | Useful for comparing AWS vs. GCP/Azure from your location, not for inter-region. |
| ThousandEyes (Cisco) | Enterprise probes, TCP + BGP | Paywalled | Quarterly PDF cloud performance report — good for order-of-magnitude citations, not real-time. |
| AWS Reachability Analyzer / Internet Monitor | Your own account traffic | Account only | Not public — AWS does not publish an inter-region latency matrix. |
Dead / dropped: cloudharmony.com (Gartner shut it down), perfops.net/latency (404), aws-latency-test.com (empty shell).
Want an independent second opinion? Run 30 Lambda-to-Lambda TCP probes for a day — about 50 lines of Python. That gives you a private dataset to compare against CloudPing's public one.
eu-west-3 (Paris, ~18 ms) qualifies as a comfortable secondary. eu-central-2, eu-south-1, eu-west-2 all work with a bit more replay lag.eu-west-3 Paris — 18 ms via terrestrial fibre through the French border. Best pick for synchronous replication.eu-central-1 Frankfurt — 31 ms, richest service catalog in EU.eusc-de-east-1 European Sovereign Cloud (Brandenburg) — 105 ms, higher because it currently routes through a limited number of PoPs.us-east-1 N. Virginia — 97 ms, the standard cross-Atlantic pivot.ap-south-1 Mumbai — 119 ms, shortest hop to Asia via the SEA-ME-WE cable path.