region·latency

Latency is distance. Light in glass covers about 100 km per millisecond of round trip, so every region has a floor you cannot buy your way under. This page measures what your machine actually gets to each cloud region, and shows how much of that is physics and how much the network added on top.

Probe origin
Seen from
detecting
Network
—
Address
—
Nearest edge
—
Link speed
Clouds loading catalog
Regions pick a cloud first
0 of 0 selected
Payload how long would this much data take to move, at each distance?
Between two regions answered from geography alone, no probes sent — a browser cannot make one datacentre ping another. Run sweep measures from you, not between them.
idle
distance ÷ speed of light everything the network added one ping †the endpoint's own server time, not distance relative comparison, not an absolute floor ·
Region Cloud Round trip ↑ Measured on the shared scale vs light Move 1 MB Distance
No measurements yet
Choose your clouds and regions above, then run a sweep to see which region is closest — in milliseconds, not kilometres.
Questions this page answers

Which cloud region is closest to me?

Select every cloud and run a sweep: the table ranks every region it can reach — 276 in the catalog — by measured round trip, so the closest cloud region, Amazon's, Microsoft's, Google's or any other, is simply the top row. Keep fastest 10 then narrows the field for a cleaner rerun.

How do I run a latency test to AWS, Azure or Google Cloud?

Pick the cloud — Amazon Web Services, Microsoft Azure, Google Cloud Platform, Oracle Cloud Infrastructure, DigitalOcean, Akamai Cloud (Linode) or Vultr — choose regions, and run a sweep. Each figure is a real HTTPS round trip from your connection to a probe host chosen for living inside its region; anycast front doors that answer from the nearest city were rejected.

Can it measure the speed between two servers?

Not directly — no browser can make one datacentre ping another. The Between two regions panel answers from geography instead: the great-circle distance, the hard floor at the speed of light in fibre, and the band where real inter-region routes land. Truly measuring server-to-server latency takes a machine inside each region.

Is this an internet speed test?

Only partly. A speed test measures bandwidth — how many bits per second your line moves. This page measures latency: how long a round trip takes, which is what makes an API, a game server or a database feel near or far. Your link speed is measured before your first sweep, but only to feed the transfer model — which is why the 1 MB column is modelled, not downloaded.

Why is my latency high on a fast connection?

Because latency is distance plus routing, and bandwidth buys neither. Fibre carries light roughly 100 km per millisecond of round trip, so 9,000 km can never cost less than about 90 ms, whatever plan you pay for. What the route adds on top — detours and queuing — is the red part of each bar. A VPN moves the starting line: everything is measured from its exit.

Where do the region lists come from?

From the providers themselves — published IP-range feeds and geofeeds, re-pulled by the page where the provider allows it. The region lists and IP ranges are never hand-typed, and probe endpoints are checked against each provider's published address space where one exists — GCP and DigitalOcean, whose feeds cannot answer that way, are confirmed by regional DNS and cross-cloud agreement. The method is documented in the colophon above.