What exactly counts as W?

Total time inside the system; queue wait plus processing. Measuring only service time understates L badly.

Why must I leave one field blank?

Three values overdetermine the law; entering all three invites inconsistent data. Two inputs uniquely determine the third.

Does this apply to async/thread-pool servers alike?

Yes; the law is measurement-model agnostic. It constrains whatever system boundary you define consistently.

Developer & Tech

Concurrency & Throughput Calculator

What this does

Apply Little's law to solve the missing variable among arrival rate, latency and concurrent requests, with capacity guidance.

Enter your details

Runs in your browser

How to use it

Using the concurrency & throughput calculator

01
Fill in any two fields
Leave exactly the one you want solved blank; the form explains itself otherwise.
02
Read the guidance row
Each result translates into a concrete pool-size or headroom statement.

Good to know

One law, three planning questions

“How many connections will 500 req/s hold open?” Solve L. “What throughput can 20 workers sustain at 200 ms?” Solve λ. “What latency holds given our pool?” Solve W. Same equation, different unknowns; which is why Little’s law anchors capacity reviews everywhere from web farms to database pools.

Averages hide the tail

The law constrains means, not percentiles. Systems average fine yet fall over during minute-scale spikes; pair this arithmetic with load testing and tail-latency budgets before treating any solved number as safe capacity.

How it's calculated

The math behind this calculator

Little's law: L = λ · W
λ req/s · W seconds · L requests in system

Little's law links three averages that must hold in any stable system: the number of requests in flight equals arrival rate times time-in-system. Enter any two values and the third follows algebraically; concurrency from traffic, sustainable throughput from pool size, or implied latency from observed load.

Assumptions & limitations

Stable system; arrivals equal completions long-run.
Averages only; bursts and tails need queueing models.
W includes queueing plus service time, end-to-end.

Worked example

Traffic arriving at 50 req/s with 120 ms average response keeps, on average, 6 requests in flight; a worker pool below 6 would queue them.

FAQ

Frequently asked questions

What exactly counts as W?: Total time inside the system; queue wait plus processing. Measuring only service time understates L badly.
Why must I leave one field blank?: Three values overdetermine the law; entering all three invites inconsistent data. Two inputs uniquely determine the third.
Does this apply to async/thread-pool servers alike?: Yes; the law is measurement-model agnostic. It constrains whatever system boundary you define consistently.

Keep exploring