Networking

Round trip time (ping)

How long a small message takes to reach another computer and come back, in milliseconds. The ping tool reports it. It decides how snappy a connection feels and limits TCP speed on long links.

Where the milliseconds come from

A round trip adds up four delays: the time the signal spends traveling, the time to put the bits on the wire, the time waiting in queues and the time each device spends processing. Distance is the part nothing can change. Light in optical fiber moves at about two thirds of its vacuum speed, which works out to about 5 microseconds per kilometer one way. A round trip doubles that, so every 62 miles (100 km) between you and the server costs about 1 ms.

New York to London is about 3,460 miles (5,570 km) in a straight line, so the floor is about 56 ms. Cables don't run straight, and Verizon's network averages about 70 ms across the Atlantic.

Typical numbers

Verizon publishes monthly round trip averages for its network. In October 2026 they were:

  • Inside Europe: about 14 ms
  • Inside North America: about 33 ms
  • Across the Atlantic: about 70 ms
  • Across the Pacific: about 112 ms

A geostationary satellite is in another class. One hop from ground to satellite and back takes about a quarter of a second, so a two way exchange takes close to half a second.

What RTT does to speed

Every new connection pays for round trips before any data moves. The TCP handshake takes one, and a TLS 1.3 handshake takes one more. At the 70 ms transatlantic average, that's 140 ms before the first request goes out.

RTT also caps a single TCP connection. With a 64 KB window on a 1 Gbps link, the ceiling is:

  • 1 ms: 524.3 Mbps
  • 10 ms: 52.4 Mbps
  • 50 ms: 10.5 Mbps
  • 100 ms: 5.2 Mbps
  • 200 ms: 2.6 Mbps

Window scaling lifts that cap, and the bandwidth-delay product tells you how big the window has to be.

Reading a ping result

The ping tool sends an ICMP echo request and waits for the reply. Mike Muuss wrote it in December 1983 and named it after sonar. It prints each reply time, then the minimum, average and maximum, plus the standard deviation.

The minimum is the closest to the pure distance cost, since queues only add delay. A big gap between minimum and maximum means queuing or a noisy link, which people call jitter. One caveat: ping time can differ from what other protocols see, because routers may treat ICMP differently from your web traffic.

Work it out yourselfOpen the Data Transfer Calculator

Sources

Back to the glossary