Technical reference
Ping technical details
This reference explains how the Ping tool runs its tests: why the pings come from Globalping probes, which options it offers, how each figure is calculated, and the limits it keeps to.
Why the pings run elsewhere
DNS Panopticon runs on Cloudflare Workers, which can make HTTP requests and open TCP connections but cannot send ICMP packets. Rather than report a different measurement under the name ping, the tool asks Globalping, a free network of community and data centre probes run by jsDelivr, to send real pings and report what they saw. Results come from those probes, not from DNS Panopticon’s own servers.
How a test runs
- The page sends the host name or address and your options to
/api/tools/ping. - The host is checked first. Private, loopback, link-local, documentation and other reserved addresses, and names such as
localhostor.local, are refused, because a public probe can’t reach them. - The API asks Globalping to start a ping measurement and returns its id.
- The page asks for the results about once a second until every probe has finished, for up to 45 seconds. Globalping asks clients not to poll faster than every 500 ms.
- Each probe resolves the host name itself, so probes may ping different addresses when a site uses a CDN, anycast or geographic DNS.
Options
| Option | Values | Notes |
|---|---|---|
| Probe locations | Worldwide, or one continent | Worldwide spreads probes in turn across North America, Europe, Asia, Oceania, South America and Africa. |
| Probes | 3, 6 or 10 | Fewer may run if a region has fewer probes online. |
| Packets each | 3, 5, 10 or 16 | Globalping allows up to 16 per probe. |
| Protocol | ICMP echo or TCP connect | TCP times the handshake to a port (443 by default) for hosts that drop ICMP. |
What the results show
- Fastest, average and slowest round-trip time for each probe, as the probe’s ping reported them.
- Lost: requests sent minus replies received.
- Jitter: the mean difference between consecutive replies, in milliseconds, computed from each probe’s reply times.
- Median average: the middle of the probes’ averages, so one distant probe doesn’t skew it.
- Latency chart: one row per probe, sorted fastest first. The bar runs from the fastest to the slowest reply, the dot is the average and each tick is one reply. Colour bands are under 50 ms, 50 to 150 ms and over 150 ms, each named in the legend and in text for screen readers.
- Raw output: what each probe’s ping command printed.
Probe times include the probe’s own network. A probe on a home connection reads slower than one in a data centre in the same city, so treat single results as indicative, not exact.
TCP connect
In TCP mode a probe opens a connection to the port and times the handshake: its SYN out and the SYN-ACK back, as RFC 9293 describes. That works for hosts that drop ICMP and shows whether the port answers at all. It does not send any application data, so a port that completes the handshake can still have a broken web server behind it.
Limits and privacy
| Limit | Value |
|---|---|
| Tests per visitor | 15 an hour, 60 a day |
| Probe network | 250 probe tests an hour without a token, 500 with one (each probe counts as one test) |
| Waiting for results | 45 seconds |
Your IP address is not sent to Globalping; requests go from DNS Panopticon’s servers. The host you enter is sent to Globalping and is visible in its measurement record. DNS Panopticon does not store results: they exist only in your browser.