Guides

Network basics: latency, hops, jitter and loss

When a connection "feels slow", it's almost always one of four things. Here's what each one means, what's normal, and how they tie together.

Latency: how long a round trip takes

Latency is the time it takes for a small message to reach another computer and for the reply to come back. The usual way to measure it is a ping: your computer sends an "are you there?" message and times how long the answer takes. That time is the round-trip time, measured in milliseconds (ms).

Your computer Server ping: "are you there?" reply: "yes" round-trip time: 23.418 ms
A ping and its reply. The time for the whole trip is the latency.

Some latency is unavoidable: signals in fibre travel at about two-thirds of the speed of light, which adds roughly 1 ms of round trip for every 100 km of distance. Everything else is time spent in equipment along the way. As a rough guide:

Where the other end isTypical round trip
Another device on your own networkunder 2 ms
A website hosted in your own country, over home broadband5–30 ms
Across the Atlantic70–100 ms

What matters most isn't the number itself but whether it's stable, and whether it's normal for that destination. 80 ms to New York is fine; 80 ms to your office router isn't. BadHop's default warning level is 100 ms, and you can set your own for each target.

Hops: the routers in between

Your traffic doesn't travel straight to the server. It's passed along a chain of routers: your own router, then your internet provider's, then possibly several other networks, until it reaches the destination. Each router along the way is a hop.

You Your router Your provider Another network Server hop 1 hop 2 hop 3 hop 4
Each router between you and the server is a hop. Real paths usually have 8 to 20.

A traceroute lists those hops in order and times each one. It works by sending messages that are allowed to travel only one hop, then two, then three, and so on; each router that has to give up on a message sends a note back, which reveals where it is and how long it took to reach. That's how you find out where a problem starts, not just that there is one. The next guide covers how to read one properly.

Jitter: how much the latency wobbles

Jitter is how much the latency changes from one ping to the next. A connection that answers in 20, 21, 20 and 22 ms has low jitter. One that answers in 18, 45, 22 and 60 ms has high jitter, even though its average might look acceptable.

Low jitter 20, 21, 20, 22, 20, 21 ms High jitter 18, 45, 22, 60, 24, 37 ms
Downloads barely notice jitter. Live audio and video do: each packet has to arrive in time to be played, so uneven arrivals mean choppy sound, frozen frames or robotic voices.

Calls and video meetings cope with a few milliseconds of jitter by holding a small buffer. Once it reaches the tens of milliseconds, the buffer can't hide it. BadHop warns when a target's jitter goes over 25 ms by default, which is deliberately early.

Packet loss: messages that never arrive

Packet loss is the share of messages that never get an answer. It's shown as a percentage: 5% loss means 5 pings in every 100 disappeared. On a healthy connection it should be zero, or very close.

Even small amounts matter. Calls break up and games stutter at a percent or two. Downloads and web pages still arrive, because lost data is re-sent, but they slow down, because the sender backs off every time it sees loss.

One thing to know: some servers and routers deliberately ignore or slow down pings when they're busy. A single host that doesn't answer isn't always broken, which is why it helps to watch several targets at once and to look at the whole route.

How they fit together

These four aren't independent. The most common story is congestion: a link somewhere is being asked to carry more than it can.

  1. Traffic starts to queue at the busy link, so latency rises.
  2. The queue grows and shrinks from moment to moment, so jitter rises.
  3. The queue fills up completely, and new arrivals are dropped: packet loss.

That's why rising latency and jitter are often an early warning of loss to come. It also helps you tell problems apart: loss without any rise in latency points less to congestion and more to something physically wrong, such as poor Wi-Fi, a damaged cable or a faulty line.

What you noticeLook atOften means
Everything feels sluggishLatencyA long or congested path
Calls are choppy, video freezesJitter, then lossCongestion, or busy Wi-Fi
Pages stall, downloads crawlLossCongestion or a faulty link
Drop-outs at particular times of dayAll three, over hoursA link that fills up at busy times

A good first test for any of these: plug your computer into the router with a cable. If the problem disappears, it's the Wi-Fi, not the internet.

See it for yourself. BadHop shows all four for every host you add, live and over hours, and traces the route so you can see where they start. Download it free: up to 3 targets, every feature.

Next: Which hop is causing packet loss?