Load Testing Vocabulary Exercises

1 / 5

A load test configuration shows: "Ramp-up period: 5 minutes, target: 500 virtual users."

What does the ramp-up period in a load test do?

The ramp-up period simulates realistic traffic growth. Without it, 500 virtual users start simultaneously — an artificial spike that doesn't reflect real traffic patterns and can trigger cold-start behaviour that wouldn't occur organically.

Phase	Description
Ramp-up	Gradually increase virtual users from 0 to the target load level
Steady-state	Hold at target load — the main measurement period
Ramp-down	Gradually reduce load to observe recovery behaviour
Think time	Pause between requests per virtual user — simulates real user behaviour

2 / 5

A load test runs for 4 hours at normal production traffic levels. At the 3-hour mark, the error rate starts increasing gradually.

What type of load test would specifically be designed to detect this kind of issue?

A soak test (endurance test) runs at expected load for many hours (4–24+). It finds issues that only emerge over time: memory leaks, connection pool leaks, disk space exhaustion from log growth, and gradual performance degradation.

Test type	What it finds	Duration
Soak / endurance	Memory/resource leaks, gradual degradation	4–24+ hours at normal load
Stress	Breaking point, failure behaviour above capacity	Short — ramp up until failure
Spike	Sudden traffic burst recovery, auto-scaling speed	Short — burst then recover

3 / 5

A load test report shows these results:
Error rate: 0.1% at 1,000 RPS | 1.2% at 2,000 RPS | 18% at 3,000 RPS

What does this non-linear pattern indicate?

Non-linear error rate increase is a saturation signature. The system handles load well up to ~2,000 RPS (the capacity knee), then degrades steeply. This data directly informs capacity planning — the service should be scaled before reaching 2,000 RPS in production.

Term	Meaning
saturation point	The load level at which the system runs out of capacity and degrades non-linearly
capacity knee	The inflection point on the error rate / latency curve where degradation accelerates
capacity planning	Using load test results to determine when to scale out before production load hits the knee

4 / 5

A performance engineer says: "We ran a spike test by injecting 10× normal traffic for 30 seconds."

What does a spike test validate?

A spike test injects a sudden, extreme traffic burst then removes it. It tests: does auto-scaling respond fast enough? Does a circuit breaker trigger? Does the system shed load gracefully or fail completely? Does it recover after the spike ends?

What a spike test validates	Key vocabulary
Auto-scaling reaction time	auto-scaling latency — how fast new instances spin up
Overload handling	circuit breaker, shed load, rate limiting
Recovery behaviour	recovery time — how long until normal after spike ends
Resilience under surge	surge — sudden demand spike event

5 / 5

A load test plan states: "300 virtual users with a think time of 2 seconds between requests."

What is "think time" in load testing?

Think time prevents unrealistic "virtual user request storms". Without think time, each of 300 virtual users sends requests as fast as possible — that's not how real users behave. Think time = pause between requests, simulating the time users spend reading content.

Term	Meaning
think time	Simulated pause between requests per virtual user — models real user behaviour
pacing	The rate at which virtual users start new iterations (related to but distinct from think time)
concurrency vs. arrival rate	Concurrency = active users at once; arrival rate = new users per second — both shaped by think time
realistic load simulation	Configuring think time and pacing to match observed real-user behaviour patterns