Donaldson R F Server Region Latency Estimator
Tell the tool where your visitors are and it ranks AWS, Google Cloud and Microsoft Azure regions by the theoretical network delay from distance alone, shown as a range.
An estimate, not a measurement. Nothing is pinged and no server is contacted. Real latency depends on the routes your traffic takes, peering between networks, congestion, your application and the visitor's own connection. Use this to shortlist regions, then measure from real users before you decide. Regions are placed at the city named by the provider; the exact datacentre is not published.
Where are your visitors?
Pick a city from the list or type coordinates such as 51.5074, -0.1278. The share is how much of your traffic comes from there (the two rows below are only an example: replace them).
Assumptions (change them if you know better)
The route factor is how much longer the real cable path is than the straight line over the Earth's surface.
Best regions for all your visitors
* The provider names only a state, country or area for this region, so the reference city is our assumption. The datacentre can be a few hundred kilometres away, which changes the estimate by a few milliseconds.
How the estimate is made
For each visitor location and each region the tool measures the great-circle distance (the shortest path over the Earth's surface, by the haversine formula with a mean radius of 6,371 km). Light in optical fibre travels at roughly two thirds of its speed in a vacuum, about 200,000 km per second. Real cables never run in a straight line, so the distance is multiplied by a route factor of 1.5 to 2 to give a typical range. The round trip is twice the one-way time:
round trip (ms) = 2 × distance (km) × route factor ÷ 200,000 × 1,000
Example: London to Frankfurt is about 640 km, so the round trip is roughly 10 to 13 ms. With several locations the tool weights each one by the share of traffic you gave and ranks the regions by the weighted middle of the range. A region is marked "about the same" when it is within 10% (or 2 ms) of the best, because ranges that close cannot be told apart by this method. Then other things should decide: price, the services you need, where the law says data must stay, and the carbon footprint.
What this leaves out. Network equipment delay, the last mile (Wi-Fi, mobile networks), the handshakes that open a secure connection (several round trips), how busy the servers are, and the fact that some routes are much better or much worse than the typical 1.5 to 2 times. A direct subsea cable can beat the lower end of the range, and traffic that has to detour through another continent can be far slower than the upper end. The "physical floor" in the tables is the same sum with a factor of 1: nothing can be faster than that.
Where each region is. AWS, Google Cloud and Microsoft Azure publish the name of each region and the place it is in, but not the address or coordinates of the datacentres. So each region is placed at the centre of the city the provider names. Where the provider names only a state or country (for example "US East (Ohio)" or "Ireland"), the tool uses the best-known city there and marks it with an asterisk: that choice is ours, not the provider's. Providers add regions often, so the lists can be out of date; check the provider's own page for new ones. For the real distance between two ordinary places see Distance Between Cities; to check where a site's address resolves use DNS Lookup or IP Geolocation.
Source / Last verified (26 September 2026): region names and locations from AWS Regions ( regions), Google Cloud regions and zones (page updated 25 September 2026; regions) and Azure regions list ( public regions). City coordinates: GeoNames (CC BY 4.0). Cloudflare and other providers are not included.