Cloudflare, Inc.
This page shows how 504 connection tests run on this site behaved on Cloudflare, Inc. (AS13335). It is a distribution of these measurements — not a verdict on the provider, and not a sample of its customers. Read the box below before quoting anything.
How to read this page
This is a distribution of measurements taken on this site, not an assessment of a provider. Everyone whose test is counted here came to a packet loss test — most of them because something was already wrong. That makes this sample systematically worse than the provider's real customer base, and no figure on this page should be read as "provider X is bad". The honest form of every number here is "of the tests run on PingPacketTest.com for this provider".
- Nobody's IP address is stored. The provider name and ASN come from a lookup done once, at the moment of the test, and the address itself is discarded — see the privacy policy.
- Ping is only shown per measurement server. A single figure averaged across measurement servers on three continents would mostly say how far away the visitor was, not what their line does.
- Medians and 90th percentiles, never averages. One half-hour test on a broken line would otherwise move the whole page.
- A provider needs at least 500 stored tests to appear here at all. Below that there is no page — not a thin one, not a placeholder. Two unlucky measurements should never be able to move a published figure.
- Window: the last 365 days. Only finished tests of at least 10 seconds with a real measurement in them.
Across all 504 tests
The first two figures are shares of tests, not of packets. A test counts as "saw loss" if a single packet failed to return, and a two-minute test at 128 packets per second sends more than fifteen thousand of them — so the longer and faster the test, the more likely it is to catch one. That makes the first column a poor measure of severity and the second one, tests above 1 % loss, the one that tracks "bad enough to feel".
Per measurement server
Ping is only meaningful next to a destination, so it is only published next to one. A visitor on this provider measuring to Singapore is measuring the distance to Singapore; the same line to a nearby server is a different number and both are correct. The gap between the median and the 90th percentile is the interesting part — a line that is fine most of the time and unusable in the bad tenth shows up as a wide gap, and an average would hide it.
| Measured against | Tests | Median ping | 90th pct ping | Median jitter | Tests that saw loss |
|---|---|---|---|---|---|
| Europe (Frankfurt) | 322 | 85 ms | 180 ms | 8 ms | 74.53 % |
| Asia (Singapore) | 130 | 35 ms | 151 ms | 3 ms | 83.08 % |
Servers with fewer than 50 tests for this provider are left out rather than shown with a number a handful of measurements could move.
Where these tests came from
A pure distribution: how the tests for this provider split across countries. There is deliberately no quality figure in this table. Cutting ping or loss by provider and country produces groups small enough that a couple of bad evenings would look like a finding.
| Country | Tests | Share of this provider's tests |
|---|---|---|
| Russia | 156 | 30.95 % |
| Indonesia | 85 | 16.87 % |
| Germany | 47 | 9.33 % |
| United States | 31 | 6.15 % |
| Iran | 25 | 4.96 % |
| Everywhere else, combined | 160 | 31.75 % |
Countries with fewer than 25 tests for this provider are combined into the last row. Cities are never published: a city row is a small enough group to start describing identifiable people.
If your own connection looks worse than this
These figures are a reference point, not a diagnosis. The useful comparison is your own line against this table, measured the same way: run a test — it starts on its own — and let it run for a few minutes rather than a few seconds, because packet loss comes in bursts and a ten-second test mostly measures luck. Every finished test can be exported as a PDF report you can send to a provider, and the latency-under-load test answers the other half of the question: whether the line falls apart the moment somebody else in the house starts a download.
What this page cannot tell you is whether the problem is the provider at all. A test measures the whole path — your Wi-Fi, your router, your provider's network, the route in between and our server. The methodology page says where the measurement stops being able to distinguish them.
Other providers with enough measurements are on the provider overview; the same tests grouped by country and by server, naming no provider, are on the packet loss statistics page. This table was rebuilt on 2026-09-07.