Latency
Definition
Latency is the time between triggering an action and the occurrence of an expected reaction.
In distributed systems, it consists of network, queue, computation, database, and rendering times. Averages often obscure slow tail requests.
Simply explained
Like the waiting time between pressing a button and visible reaction.
Why it matters in practice
Relevant for websites, APIs, voicebots, and AI assistants.
Example from business
Possible example: A team separately measures the wait time until the first text snippet and until the complete answer in an assistant.
Limits and pitfalls
- Only measure average; mix client and server latency; do not separate warm/cold starts.
Origin & history
Latency is a fundamental concept in networking and performance engineering; SRE practice often considers percentiles rather than just averages.
Use in mid-sized companies
- Possible application in SMEs: measure p50/p95/p99 and user-relevant partial times.
Frequently asked questions
Is latency the same as throughput?
No. Why p95? It better shows slow user experiences than just the average.
Sources
- Monitoring Distributed Systems · Google SRE Book, Monitoring Distributed Systems
Sources checked on 10.10.2026.
Editorially reviewed on: