Latency test
Before you choose where to run a server, measure how far each place is from you. Your browser sends a few small requests to each provider's published test endpoint and times them; nothing goes through ANAROQ, and ANAROQ is not one of the places.
What leaves your browser
When you press Run the test, your browser sends 60 small requests (6 per endpoint, one after the other) directly to the 10 endpoints listed below, run by Cloudflare and Vultr. Each request carries nothing about you, but like any request, the provider sees your IP address and browser. Nothing is sent before you press the button.
The results stay in this page. They are not sent to ANAROQ, not stored and not sent to analytics; only sharing the report puts the numbers in the link, after # (browsers never send that part to a server).
The endpoints and why they may be used
- Nearest edge · Cloudflare · speed.cloudflare.com
Permission: the provider publishes a client for measuring against it (github.com, checked 18 Oct 2026) - Tel Aviv · Vultr · tlv-il-ping.vultr.com
Permission: the provider publishes it as a public test endpoint (tlv-il-ping.vultr.com, checked 18 Oct 2026) - Frankfurt · Vultr · fra-de-ping.vultr.com
Permission: the provider publishes it as a public test endpoint (fra-de-ping.vultr.com, checked 18 Oct 2026) - Amsterdam · Vultr · ams-nl-ping.vultr.com
Permission: the provider publishes it as a public test endpoint (ams-nl-ping.vultr.com, checked 18 Oct 2026) - London · Vultr · lon-gb-ping.vultr.com
Permission: the provider publishes it as a public test endpoint (lon-gb-ping.vultr.com, checked 18 Oct 2026) - Paris · Vultr · par-fr-ping.vultr.com
Permission: the provider publishes it as a public test endpoint (par-fr-ping.vultr.com, checked 18 Oct 2026) - Madrid · Vultr · mad-es-ping.vultr.com
Permission: the provider publishes it as a public test endpoint (mad-es-ping.vultr.com, checked 18 Oct 2026) - Milan · Vultr · mxp-it-ping.vultr.com
Permission: the provider publishes it as a public test endpoint (mxp-it-ping.vultr.com, checked 18 Oct 2026) - Stockholm · Vultr · sto-se-ping.vultr.com
Permission: the provider publishes it as a public test endpoint (sto-se-ping.vultr.com, checked 18 Oct 2026) - Warsaw · Vultr · waw-pl-ping.vultr.com
Permission: the provider publishes it as a public test endpoint (waw-pl-ping.vultr.com, checked 18 Oct 2026)
Middle East outside Israel: no provider we checked publishes a test endpoint there that a browser may use, so there is none in the list.
What it does
It measures the round trip from your browser to test endpoints that providers publish for this use, in Israel and in Europe, plus Cloudflare's nearest edge as a baseline for your own connection. For each place it takes several samples, leaves out the first one (it also pays for DNS, the connection and TLS) and shows the median, what a typical request costs, and p90, how slow the slow ones get.
How it works
- Read the notice and the list of endpoints: each one links to the page where its provider allows this use.
- Press Run the test. The endpoints are measured one after the other, a few seconds in all.
- Read the results by area. A lower median means a shorter path from you; a p90 far above the median means an unsteady one.
- Share the report if you like: the link holds the numbers only, not where you are.
How the numbers are made
- Samples
- 6 requests per endpoint, in a row. Each has a fresh cache-busting parameter and is sent with no cookies and no cache, so it goes over the network every time.
- Warm-up
- The first request also looks up the name and opens the connection (TCP and TLS), so it is not counted.
- Median and p90
- Of the counted samples: the middle value, and the value 90% of the samples do not exceed (nearest rank). A request that fails or does not answer in 4 seconds is not counted.
- Timing
- The browser's own Resource Timing for each request, from the start of the fetch to its last byte; the page's clock when the browser gives none.
Example
For illustration (made-up samples, not a measurement): a browser in Tel Aviv on a home fibre line. The first sample of each series is the warm-up, and one Stockholm request did not answer. These are the medians and p90 the test computes from those samples:
Israel
| Location | Provider | Median (ms) | p90 (ms) | Samples |
|---|---|---|---|---|
| Tel Aviv | Vultr | 5.1 | 12.7 | 5 |
Europe
| Location | Provider | Median (ms) | p90 (ms) | Samples |
|---|---|---|---|---|
| Frankfurt | Vultr | 58.8 | 61.2 | 5 |
| Amsterdam | Vultr | 65.9 | 71.5 | 5 |
| Stockholm | Vultr | 79.2 | 80.1 | 4 |
Nearest edge (anycast)
| Location | Provider | Median (ms) | p90 (ms) | Samples |
|---|---|---|---|---|
| Nearest edge | Cloudflare | 7.0 | 9.9 | 5 |
Limits
- A browser measurement includes your own Wi-Fi, router and browser, and a busy device or another download skews it. Compare places measured in the same run.
- It is an HTTP round trip on an open connection, close to but not the same as ping (ICMP), and not what a server you rent there will answer: that depends on the server and its network.
- One provider's endpoint stands for a location; other providers in the same city may be on different routes.
- Results change during the day. Run it more than once before you decide.
- The list has only endpoints a provider publishes for this use; a location with none is not in it. ANAROQ has no region and is never in it.
Methodology: how the numbers are worked out
Sources
Related
Next step
A free public tool by ANAROQ, a company building compute infrastructure. What ANAROQ is building