Realistic simulation.
WHAT MAKES A LOAD TEST VALID
Realistic user behaviour Realistic data volume Realistic mix of operations Realistic distribution over time
WHAT MAKES ONE INVALID
Every user performing the same action Requests with no pauses A tiny dataset Caches warmed by repetition
WHY THAT LAST POINT MATTERS
Repeated identical requests hit caches, producing results unrelated to real load.
WHAT TO VARY
Data used by each virtual user.
WHAT TO MODEL
Think time between actions The proportion of each operation Peak concurrency
HOW TO ESTABLISH THOSE
Production traffic data, where available.
WHAT TO RAMP
Load gradually, observing where degradation begins.
WHY GRADUALLY
An immediate full load reveals the breaking point but not the curve.
WHAT THE CURVE SHOWS
The point where response time rises sharply while throughput stops increasing.
WHAT THAT POINT IS
The practical capacity.
WHAT TO RUN THE TEST FROM
Enough capacity that the test tool is not the bottleneck.
WHAT TO VERIFY
That the tool itself is not saturated.
WHAT TO TEST AGAINST
An environment resembling production, in configuration and data volume.
