What the four figures mean
| Figure | What it tells you |
|---|---|
| Median | The typical round trip: half the samples were faster, half slower. One slow sample does not move it. |
| Minimum | The fastest sample — what the route can do when nothing on the way is queueing. |
| 90th percentile | Nine in ten samples were this fast or faster. A big gap to the median means regular slow moments. |
| Jitter | The median change from one sample to the next. Calls and games feel it more than a slightly higher ping. |
There is no average on purpose. Round trips are not spread evenly: most are close to the minimum and a few are much slower, when a Wi-Fi frame has to be resent or another download fills a queue. An average lets those few decide the figure; the median and the 90th percentile show both halves of the story.
Reading the chart
- Even, low columns — a quiet line. The median sits close to the minimum and jitter is a fraction of a millisecond to a few milliseconds.
- Single tall columns — short queues somewhere on the way, very often the local Wi-Fi. Run the test again over a cable: if they disappear, the radio link was the cause.
- Many tall columns, or a rising staircase — the line is busy. Something on the same connection is uploading or downloading; a full upload in particular delays every other packet (bufferbloat).
- A high minimum — distance or routing, not your line: no setting on your side lowers it, only a shorter route or a closer server.
Ping our test server from your own machine
The command line can do what the browser cannot: send real ICMP pings. Our test server answers them on both addresses.
ping -c 20 45.153.34.90
ping -6 -c 20 2a0f:ca80:b00b:d657::5ping -n 20 45.153.34.90The last lines of the Linux and macOS output carry the same figures in another form: min/avg/max/mdev, where mdev is the spread of the samples. For a route that drops or delays packets at one point, MTR shows every hop on the way — the looking glass runs it from our side.
Where the test runs
The samples go to our test server on the TechTies network (AS197170) in Amsterdam, and the ping back comes from it. There is no test server in Frankfurt yet; for most of Europe the two cities are close neighbours, and Frankfurt or Amsterdam shows how to compare them for your own users.
Frequently asked
Can a website send a real ping?
No. Browsers cannot send the ICMP echo requests the ping command uses. This test times HTTPS requests instead: two warm-up requests open the connection, and each sample after that is one small request on the open connection, timed by the browser from sending it to the first byte of the answer — one network round trip plus well under a millisecond on the server. The second part of the page sends real ICMP pings, from our server in Amsterdam to your address.
What is a good ping?
Distance sets the floor. Light in optical fibre covers about 200 km per millisecond, so a round trip to a place 1,000 km away cannot take much less than 10 ms, and real routes are longer than a straight line. On top of that come your own network — Wi-Fi adds a few milliseconds and some jitter, mobile networks more — and your provider's route. A remote shell or desktop feels immediate well below 50 ms; fast online games are where a few milliseconds start to matter.
What is jitter, and how is it worked out here?
Jitter is how much the round-trip time changes from one packet to the next. This test takes the difference between each sample and the one before it and reports the median of those differences, so one stray spike does not decide the figure. Voice and video calls notice jitter more than a steady, slightly higher ping, because their buffers have to absorb the variation.
Why does the ping from Amsterdam to my address get no reply?
Most home routers and many firewalls drop ping requests that come from the internet, so 100% loss to a home connection is normal and says nothing about its quality. If your router does answer, the figures are the round trip from our network to it, measured exactly as the ping command does. A traceroute on the looking glass shows the route up to the last router that answers.
Is my measurement stored?
No. The samples are timed by your browser and shown only to you; our test server answers each one with an empty response and does not log successful requests. The ping from Amsterdam is logged like every looking glass test, with your address cut down to its network (the first three parts of an IPv4 address).