Latency

Latency is the time it takes for data to travel from source to destination, or more broadly the delay between an action and its observable effect. In networking it is often measured as round-trip time, the interval between sending a request and receiving the start of a response.

How It Works

Network latency comes from several contributors. Propagation delay is bounded by the speed of light over distance, so a request crossing continents has an unavoidable floor. Transmission, queuing, and processing delays add to it at each hop, switch, and server along the path. Protocol behavior matters too: TCP and TLS handshakes add round trips before any useful data flows, which is why HTTP/3 works to reduce setup cost.

Latency is best understood as a distribution, not a single number. Teams track percentiles such as p50, p95, and p99, because tail latency, the slowest requests, often shapes user experience and triggers timeouts more than the average does.

Why It Matters

Low latency makes applications feel responsive. It is critical for interactive systems, real-time communication, trading, and gaming, where small delays are perceptible or costly. High latency frustrates users and can cascade into retries and overload.

Latency and throughput are distinct: a link can have high throughput yet poor latency, or low latency yet limited throughput. Common techniques to reduce latency include placing content closer to users with a CDN, caching, connection reuse, reducing round trips, and choosing efficient protocols. Distance and the speed of light set hard limits that no amount of bandwidth can overcome.

Related Terms

Latency is distinct from throughput, is shaped by TCP and UDP behavior, and is reduced by serving content from a nearby CDN.