Stress Testing

A load test proves you can handle the expected. A stress test tells you what happens when the unexpected arrives — where the ceiling is, and how the system behaves when it hits it.

What it is

Stress testing deliberately pushes load beyond expected peaks until the system degrades or fails. The goal is not a pass/fail verdict but characterisation: the maximum sustainable throughput, the first resource to saturate, the failure mode, and the recovery behaviour once load subsides.

Why failure mode matters

Two systems with the same capacity ceiling can behave very differently at that ceiling. One sheds load gracefully — fast errors, circuit breakers opening, queues bounded. The other degrades catastrophically: latencies climb into the tens of seconds, retries amplify the load, queues grow unbounded and the system enters a death spiral that persists even after traffic drops. Knowing which one you have is the difference between a slow Saturday and a front-page outage.

Our approach

We ramp load in steps — typically 25% increments above the validated load-test plateau — holding each step long enough to observe steady-state behaviour. At each step we record the knee points: where latency begins its non-linear climb (see queueing theory), where errors begin, where throughput plateaus and then drops. After failure, we cut load and measure time-to-recovery without intervention.

Step profile: stress test
  step 1 : 100% of peak  (validated baseline)
  step 2 : 125%   step 3 : 150%   step 4 : 200%
  step 5 : +25% increments until failure
  observe: first saturated resource, failure mode,
           recovery time after load removal

Outputs you can act on

A measured capacity ceiling with the limiting resource identified; verification (or refutation) that degradation is graceful; evidence for capacity planning and auto-scaling thresholds; and specific hardening recommendations — timeouts, retry budgets, circuit breakers, load shedding.