How long is this line?#
SLOP2710 · Week 3
What you can see from where you stand#
Ahead of you — count one switchback, multiply. Say 240.
Speed — pick a fixed point, count crossings for two minutes. Say 80, so 40 a minute.
No access. No data request. Nobody to ask.
240 ÷ 40#
= 6 minutes#
L = λW#
average number in · arrival rate · average time each stays
What it does not care about#
- any arrival pattern
- any service time distribution
- any number of servers
- any discipline whatsoever
- L
- —
- λ
- —
- L ÷ λ
- —
- W
- —
- W90
- —
Same area, two cuts#
Count across — how many are here right now? Average it. That is L.
Follow one through — how long did this one stay? Average it. That is W.
Both are measuring person-minutes.
The proof, in four lines#
- Window of length T. N people pass through.
- Total time inside = A. So W = A/N.
- Same A, integrated over the window: L = A/T.
- And λ = N/T.
L/λ = (A/T)÷(N/T) = A/N = W
Now the dangerous part#
Where is the box?#
Three boxes, one airport#
| Box | W |
|---|---|
| The arch | 0.5 min |
| The hall | 60 min |
| Kerb to gate | 85 min |
All correct. All publishable as “the average wait”.
State the box#
before you state the number#
What it is for#
Estimate a wait from outside.
Size something you have not built.
Catch a lie — 500 people, 20 a minute, and they claim 10 minutes.
Sanity-check a promise — halve W with λ fixed, and L must halve. Which half of these people is leaving?
It is not really about queues#
- Factory: WIP = throughput × cycle time
- Kanban: lead time = WIP ÷ throughput
- Warehouse: inventory turns
- Support desk: 300 open ÷ 60 a day = 5 days
- Hospital: beds = admissions × length of stay
Three ways to get it wrong#
It was filling or emptying. Steady state, or nothing.
Different windows. L at 12:15 and λ across three hours are two systems.
People left without being served. They were in L. Were they in λ?