Glossary · Performance

What is Latency?

Latency is how long data takes to get there and back. It is measured in milliseconds and it is not the same thing as bandwidth: a connection can be very fast and still feel sluggish if latency is high, because every interaction waits for a response.

Term
Latency
Also known as
Delay, round-trip time
Category
Performance
In short
The time it takes a packet to travel from one point to an…

In detail

The floor set by physics

Light travels through fiber at roughly two-thirds the speed of light in vacuum, which works out to about 5 microseconds per kilometre in each direction — near enough 1 millisecond of round trip for every 100 kilometres of path. No provider can beat that, and any claim that implies otherwise is wrong. What a provider controls is how much is added on top.

Where the extra milliseconds come from

Every routed hop, every queue, and every piece of processing adds delay. A congested link adds the most, because packets wait. This is why a dedicated circuit with few hops behaves consistently and a path across several congested networks does not, even when both are nominally fast.

Path length is not map distance

Fiber follows rights of way, not straight lines. Two cities 200 kilometres apart may be joined by a 300-kilometre fiber route, and a protected service may take a much longer path after a switch. When latency genuinely matters, ask about the route and about what happens to latency during protection switching.

Measuring it honestly

A ping is a weak proxy. Routers frequently deprioritise ICMP, so a ping figure can look worse than the traffic that matters or hide a problem the application feels. Measure with the protocol you actually run, over the path you actually use, and look at the distribution rather than the average: the ninety-fifth percentile tells you what users experience on a bad afternoon, and the mean does not.

Why it matters

Latency governs how responsive interactive applications feel and sets hard limits for synchronous storage replication and real-time control. For most office workloads consistency matters more than the absolute number; for trading, industrial control, and replication, the absolute number is a design constraint.

Where you will see it

  • A hosted application that feels sluggish despite ample bandwidth
  • Synchronous replication limited by the distance between data centres
  • A voice path that degrades when latency and jitter rise together

More performance terminology

Working through a connectivity decision?

Describe the requirement and an engineer will tell you which service actually fits — without a rate card or a sales script in the way.

Talk to a Network Specialist

Service availability depends on location, network proximity, capacity, and engineering review. Share an address and we will confirm what can be delivered there.