Communication / 2 / 6
Message queues
A shop sends an order email through a provider that takes 100 messages a second, and a sale brings 600 orders a second. How a queue absorbs the spike, and what you must set on it: leases, acks, backoff, a dead-letter queue and backpressure.
- 14 min
- Intermediate
- 12 steps
- Order p95
- 950 ms
- Backlog
- 0
- Email delay
- 1 s
- Lost emails
- 0 /h
- request
- reply
- redelivery
- duplicate
- lost
- call: the reply returns on the same line
Numbers are a model scenario, not measurements
The real request
Shop API → Mail providerRequest
POST /v1/send HTTP/1.1
Host: mail.example.com
Authorization1: hidden
{
"to": "u_42@example.com",
"template": 2"order_paid",
"order_id": "ord_5521"
}
Response
HTTP/1.1 202 Accepted
X-Response-Time3: 820 ms
Authorization· hidden — The shop's API key at the provider. Hidden here.template— Which email to send. The provider fills it with the order data.X-Response-Time— 820 ms of every order request is spent waiting for someone else.
What happens if…?
Put in your own numbers: the result and the formula update at once.
Jobs per second put into the queue.
Workers taking jobs from the queue in parallel.
How long one worker needs to send one email.
Result
- Utilisation ρ
- 0.80
- Chance to wait
- 80%
- Mean wait in queue
- 80ms
- Total (wait + work)
- 100ms
- Mean jobs waiting
- 3.2
Working
μ = 1000 / 20 ms = 50.0/s
ρ = λ / (c × μ) = 40 / (1 × 50.0) = 0.80
P(wait) = ErlangC(c = 1, λ/μ = 0.80) = 0.800
Wq = P(wait) / (c × μ − λ) = 0.800 / 10.0 = 80.0 msA simplified teaching model: it shows direction and scale, not a benchmark.
Build it yourselfCheat sheetCompare lessons
Course: 74 lessons
01 Foundations
02 Scaling
03 Data & storage
04 Communication
05 Reliability
06 Architecture
07 Security
08 Shipping
09 Whole systems
10 Business processes
11 Case studies