Before Peak Traffic

Peak traffic changes load shape and downstream demand. Test the event’s conditions, not average load.

Questions worth answering

  • Which transaction must still complete under peak?
  • What traffic mix, data shape, and duration are representative?
  • Which downstream limit can dominate the flow?

The outcome to protect

Protect accepted transactions from queue saturation, downstream throttling, and correctness failures.

How we investigate

Test bursts and sustained load, queue growth, autoscaling delay, dependency throttling, graceful degradation, and correctness within an agreed scenario.

Decisions & outcomes

  • A scenario-specific capacity envelope
  • Known limiting dependencies
  • Prioritized validation and engineering actions

Fast requests can still leave reservations behind

A fictional scenario, not a customer result.

The question to resolveWill a promotion preserve inventory reservations when the reservation consumer falls behind?

  1. Promotion traffic
  2. Reservation queue
  3. Inventory consumer
  4. Reservation confirmed

Observed in the exercise

The agreed staging burst left a reservation backlog after request latency returned to normal. Queue drain and reservation correctness need separate acceptance criteria.

Evidence basis
Tested
Verification result
Partially verified
Freshness
Fresh

Uncovered dependency

The supplier inventory feed was simulated. Its production response during the promotion is unknown.

Evidence basis
Unknown
Verification result
Unverified
Freshness
Unknown

The resulting decision

Define a safe backlog boundary and admission-control response; verify drain and reservation expiry with representative dependencies before the promotion.

Scope of this example

This fictional exercise reports no traffic capacity or customer result. The actual load shape and acceptable outcomes must be agreed for the selected flow.

Choose one critical flow.

Tell us what must work and what you need to know. We agree the scope before delivery.