The ping command in Linux
ping sends ICMP echo requests to a host and reports each reply with its round-trip time. It is the first question of all network debugging — "can I reach it at all?" — and its output encodes reachability, latency, jitter and packet loss in a form every engineer learns to read at a glance.
How ping works
ping speaks ICMP, the control protocol that lives beside TCP and UDP: it sends an echo-request packet carrying a sequence number and a timestamp, and the target's kernel — not any application — answers with an echo-reply. The round-trip time is computed from the embedded timestamp, and gaps in the sequence numbers are how ping detects loss. Because ICMP requires raw sockets, ping is granted special capabilities at install time; it is one of the few unprivileged commands allowed to craft packets.
Two fields turn ping into a measuring instrument. TTL (time-to-live) is decremented by every router; the value in replies hints at hop count, and a TTL that changes between replies betrays route flapping. Payload size (-s) probes MTU problems: if 1400-byte pings pass and 1500-byte pings vanish, something on the path fragments badly — the classic diagnosis for "SSH works but transfers stall".
Syntax
ping [OPTIONS] HOSTCommon options
| Option | What it does |
|---|---|
-c N | Send N packets and stop (otherwise ping runs until Ctrl+C). |
-i SEC | Interval between packets (default 1 second). |
-W SEC | Timeout waiting for each reply. |
-s SIZE | Payload size in bytes — for testing MTU issues. |
-4 / -6 | Force IPv4 or IPv6. |
-q | Quiet: only the summary line. |
How to use ping: examples
$ ping -c 4 google.comFour probes and a summary: reachability, min/avg/max latency and loss.
$ ping -c 1 -W 2 10.0.0.5 && echo up || echo downA one-shot liveness test with a 2-second timeout — the building block of countless monitoring scripts.
$ ping -c 100 -i 0.2 server | tail -2A quick loss/jitter measurement: 100 fast probes, read only the statistics.
$ ping -6 example.comVerify IPv6 connectivity specifically.
Real-world use cases for ping
Bisecting an outage
The site is down: ping the server's IP, then the gateway, then 8.8.8.8, then a domain name. The first failure locates the layer — machine, LAN, internet or DNS — in under a minute, and determines who you call.
Watching a reboot come back
ping server in one terminal while the machine restarts: replies stop, then resume the moment the network stack is up. The cheapest possible "is it back yet?" indicator during maintenance windows.
Pro tips and common mistakes
- Always ping both the name and the IP — the pair distinguishes DNS problems from connectivity problems instantly.
- No reply does not mean down: many hosts firewall ICMP. Confirm with a TCP probe (curl, ss on the far side) before declaring death.
- Rising latency with intact delivery is congestion or Wi-Fi trouble; loss is a different, worse disease. Read both numbers.
Frequently asked questions about ping
ping works by IP but not by name — what does that mean?
Routing is fine; DNS resolution is broken. Check /etc/resolv.conf and test the resolver directly with dig.
What is a good ping time?
Same LAN: under 1 ms. Same country: 5–30 ms. Cross-continent: 80–200 ms. Consistency matters as much as the average — variance (jitter) is what breaks calls and games.
Why does a host not answer ping but serve web pages?
Many firewalls drop ICMP while allowing TCP. No ping reply proves little; a TCP check (curl, nc) is the more honest liveness test for services.
Related commands
Download files, test APIs, send POST requests with JSON, and inspect headers.
ipThe modern tool for addresses, links and routes — replacing ifconfig and route.
digAsk DNS questions directly: A, MX, TXT records, and which server answered.