Little's Law

Items in a system = arrival rate × time each spends in it.

Backend Engineer · Senior Data Engineer · Senior

Also known as: Little's law, littles law, L = λW

Little’s law is a simple, remarkably general relationship from queueing theory:

L = λ × W
L = average number of items in the system
λ = average arrival rate (items per second)
W = average time each item spends in the system

It says the number of things “in flight” equals how fast they arrive times how long each stays. It holds for any stable system — queues, servers, databases, connection pools — regardless of the details inside.

Why it matters: it lets you reason about capacity without simulating. If your service receives 1,000 requests per second and each takes 100 ms, then on average 1,000 × 0.1 = 100 requests are in flight at once. That’s 100 concurrent requests your system must be able to hold — threads, connections, memory — at that throughput and latency. If latency rises to 1 second at the same arrival rate, in-flight requests jump to 1,000, and the system may fall over trying to hold them.

It also works in reverse: if you know the concurrency your system can sustain and the arrival rate, it tells you the latency you can expect — or diagnose that a growing number of in-flight items means latency is climbing.

The classic mistakes:

Ignoring that it applies to stable systems. If arrivals exceed capacity, the queue grows without bound and “average” items isn’t meaningful — the system is diverging, not in steady state. Little’s law describes the steady state.
Forgetting concurrency costs resources. Every in-flight item holds memory, a connection, or a thread. High latency at high throughput means many concurrent items — which is often the real capacity limit, not CPU.
Using it on a component in isolation. Scope matters: L, λ and W must refer to the same boundary (the whole service, or one queue).
Treating latency as independent of load. As load rises, W usually rises, so L rises faster than λ. The relationship exposes that feedback.
Assuming it prescribes a fix. Little’s law describes; it tells you in-flight count, not whether to add workers or reduce latency.

Little’s law is a powerful sanity check for performance and capacity: it connects the three numbers you care about — throughput, latency and concurrency — so that improving one often means paying in another. It’s why capping concurrency and reducing latency matter for backpressure, and it belongs in any capacity planning discussion.