Latency

What is it?

Network latency is the time required for data or a request to travel between points. It may be measured one-way or as a round-trip time, and it includes propagation, transmission, processing, and queueing delays.

Why does it matter?

Latency strongly affects interactive APIs, remote terminals, gaming, and distributed transactions. A high-bandwidth link cannot make propagation or protocol round trips disappear.

Mental model

Split total delay into components: serialization time to put bits onto a link, propagation time for the signal to travel, processing time at devices, and queueing time waiting behind other traffic.

How it works

• Propagation delay depends on physical distance and signal speed in the medium.
• Transmission/serialization delay depends on packet size and link rate.
• Queueing delay grows under contention and may dominate under congestion.
• Round-trip time includes the forward and return paths, which may not be symmetric.
• Application protocols that require multiple sequential round trips can multiply user-visible latency.

Steps

• Measure a baseline RTT to separate network delay from application processing time as far as tools allow.
• Check packet size, path distance, loss, and queueing metrics.
• Trace how many sequential request-response exchanges the application requires.
• Reduce unnecessary round trips or place compute/data closer when the workload justifies it.

Real-world usage

• Profiling interactive applications and database calls across regions.
• Comparing bandwidth improvements with protocol and geographic improvements.
• Diagnosing queueing spikes and tail latency.

Common mistakes

• Confusing RTT with one-way latency.
• Assuming low ping guarantees high application throughput.
• Ignoring p95/p99 latency when averages look healthy.

Interview questions

Prerequisites

• Bandwidth

Continue learning

• Transmission Media