Latency
Latency is the delay between sending a request and receiving a response — the time it takes for data to travel from one point to another on a network.
Low latency means fast response; high latency means noticeable delay. It is the reason a video call might feel sluggish or a game might lag, even when your internet speed is otherwise fast.
Latency, explained at three levels
At home
Latency is the slight delay between doing something online and seeing the result — like clicking a link and waiting a moment before the page starts loading, or speaking on a video call and the other person hearing you a fraction of a second later. Even if your internet speed is fast, high latency causes things to feel laggy and unresponsive. It is measured in milliseconds (ms), and lower is better.
At work
Latency affects every interaction employees have with network-based systems — from how responsive a cloud application feels to the quality of video and voice calls. High latency makes applications feel sluggish, causes voice calls to have awkward overlapping pauses, and reduces productivity. When evaluating cloud services or planning network infrastructure, latency between the users and the service should be measured and factored into the decision, especially for real-time or interactive applications.
At home
Latency has several components: propagation delay (the time light or electricity takes to physically travel through cables — roughly 5 ms per 1,000 km of fibre), processing delay (time routers and switches take to handle each packet), queuing delay (time spent waiting in buffers during network congestion), and serialisation delay (time to push all the bits of a packet onto the wire). Round-trip latency (RTT) captures the total time for a request to go out and the response to come back. For most web interactions, round-trip latency under 50 ms feels responsive; above 200 ms, users notice noticeable lag.
At work
Application performance is often more sensitive to latency than to raw bandwidth. A 1 Gbps connection with 200 ms latency can feel slower for interactive tasks than a 100 Mbps connection with 10 ms latency. This is because protocols like TCP require round-trip acknowledgements — each additional millisecond of latency is multiplied across every exchange. Techniques to mitigate latency include placing servers closer to users (CDNs, edge compute), reducing the number of network round trips (HTTP/2 multiplexing, connection pooling), and using UDP-based protocols (QUIC) for latency-sensitive applications.
At home
Tail latency — the latency experienced by the slowest requests (99th or 99.9th percentile) — is a critical metric in distributed systems because even a small percentage of slow responses affects a significant number of users at scale. Jitter (variation in latency over time) is particularly damaging to real-time audio and video, which require consistent packet delivery timing; jitter buffers compensate but add their own delay. Geographical latency has a hard physical floor: the speed of light in fibre is approximately 200,000 km/s, meaning a round trip between London and Sydney takes at least 70 ms in theory and typically 250-300 ms in practice due to routing and processing.
At work
Latency optimization at scale involves content and compute placement strategies (multi-region deployment, Anycast routing), protocol-level improvements (TLS 1.3 0-RTT resumption, TCP Fast Open, QUIC), kernel-level tuning (TCP buffer sizes, congestion control algorithms like BBR), and hardware-level solutions (kernel bypass with DPDK or XDP for ultra-low-latency applications in financial trading or telecommunications). Latency budgets should be decomposed across each component in the request path (DNS resolution, TCP handshake, TLS handshake, server processing, data transfer) to identify optimisation opportunities. Observability platforms with distributed tracing (OpenTelemetry, Jaeger) enable per-hop latency analysis across microservices architectures.
Real-world examples
At home
When Josh plays an online game, he notices his character responds instantly to his controls when latency is low (around 20 ms). But when latency spikes to 200 ms — maybe because someone in his household started a large download — there is a frustrating delay between pressing a button and seeing the action happen on screen.
At work
When Sophie's company evaluates two cloud providers for hosting their customer-facing web application, the deciding factor is latency. Provider A hosts in a data centre 50 ms from most of their users, while Provider B hosts 150 ms away. Testing shows that the 100 ms difference makes pages load noticeably slower on Provider B, directly impacting customer satisfaction and conversion rates.
Common confusions
- It's tempting to think faster internet speed means lower latency. Bandwidth and latency are independent; a fast connection can still have high latency if the data has to travel a long distance.
- Some people are convinced that latency only matters for online gaming. It affects video calls, cloud applications, remote desktops, and any real-time interaction.
- A lot of people believe latency is entirely fixable with better equipment. The speed of light through fibre sets a physical limit on how low latency can go over long distances.
For analysts
When a stakeholder says the system 'needs to be responsive,' translate that into a measurable latency requirement. Specify acceptable response times at the application level — for example, 'the dashboard must load in under 2 seconds for 95% of users in the UK.' Then work with the technical team to determine whether that is achievable given the network path, hosting location, and application architecture. Vague performance requirements lead to disappointment.
History & origin
Latency has been a consideration in telecommunications since the telegraph era, but it became a widely discussed term in computing with the growth of the internet in the 1990s. As real-time applications like online gaming, video calling, and financial trading became common, latency moved from a technical metric to a user-experience concern that everyone notices. The push for lower latency has driven innovations from CDNs to edge computing to 5G networks.