DNS resolution time
Enter a domain name. It is resolved once in your browser, and the time the DNS lookup took is shown in milliseconds. Nothing about the site behind it is fetched or measured.
Measurement details
How the number is obtained
A browser cannot read DNS timings directly, so the page makes the lookup happen in isolation instead. Chrome resolves hostnames through a pool of 64 worker slots; the page fills 63 of them with lookups for names on a zone it controls, where the server answers only after a delay. Once the pool is saturated, the domain under test is started with exactly one slot free, so its lookup runs alone. Two marker lookups queued on either side of it provide timestamps, and their own cost is measured and subtracted in the same run.
Only the DNS lookup is timed. TCP connect, TLS and page loading are not included.
Accuracy
The instrument's own noise is about ±2 ms; below ~5 ms the result is at its floor. Your resolver adds its own variance on top — for a name that is not cached, a single sample can be off by tens of milliseconds, because the answer depends on which recursive resolver instance answers and how it retries. The median of three samples is the stable number.
Because the method deliberately congests your resolver with 63 lookups, a name that is not already cached can read somewhat slower than during quiet browsing, while a cached name is unaffected.
The 64-slot pool is a Chromium implementation detail, so use Chrome, Edge, Brave or another Chromium browser. Re-measuring the same name within a minute returns ~0 ms because the browser caches the answer.
Technical version — all parameters, per-sample numbers, queue calibration.