In 30 seconds
- Measures speed, capacity and stability under load.
- Types: load, stress, soak and spike tests.
- Start early; performance is hard to fix late.
Main types
| Type | Question |
|---|---|
| Load | Does it cope with expected traffic? |
| Stress | Where does it break? |
| Soak | Does it stay stable for hours or days? |
| Spike | Can it handle sudden surges? |
What to measure
- Response time, especially slow outliers (95th or 99th percentile)
- Throughput: requests handled per second
- Error rate under load
- Resource use: CPU, memory, database
In agile teams
- Put performance targets in acceptance criteria where it matters
- Run small performance checks in the pipeline
- Run bigger tests before major releases
Myth vs reality
Myth · Performance testing happens at the end.
Reality · Late discovery means expensive fixes.
Myth · Averages tell the story.
Reality · Slow outliers are what users notice.
Myth · Test environments match production.
Reality · Differences must be understood and accounted for.
FAQ
What tools are common?
Tools such as JMeter, Gatling, k6 and Locust.
How do we set targets?
From user expectations and current production data.
Who owns performance?
The whole team, with specialists to guide.