Catalog
method#Quality Assurance#Reliability#DevOps#Observability

Stress Testing

Targeted testing method to verify system stability under extreme load and resource exhaustion.

Stress testing is a targeted performance testing method that evaluates system behavior under extreme load and resource limits.
Established
Medium

Classification

  • Medium
  • Technical
  • Architectural
  • Intermediate

Technical context

Monitoring tools (Prometheus, Grafana)Load testing tools (k6, JMeter)CI/CD pipelines to automate test runs

Principles & goals

Generate realistic load profiles, not only synthetic spikes.Define measurable objectives (SLO/SLA) as success criteria.Integrate observability: metrics, traces and logs must be enabled.
Run
Team, Domain

Use cases & scenarios

Compromises

  • Wrong conclusions due to unrealistic load profiles.
  • Impact on production environments from uncontrolled tests.
  • Over-optimization for synthetic tests instead of real usage.
  • Integrate automated, reproducible tests into CI
  • Ensure observability data is fully available during tests
  • Increase tests incrementally instead of starting at max load

I/O & resources

  • Load profiles, test scripts and anonymized user data
  • Representative test environment or production-like staging
  • Monitoring and alerting configuration
  • Report with thresholds, bottlenecks and remediation recommendations
  • Metric and log archives for root-cause analysis
  • Updated capacity and runbooks

Description

Stress testing is a targeted performance testing method that evaluates system behavior under extreme load and resource limits. It identifies breaking points, resource exhaustion modes, and degradation patterns to guide capacity planning and resilience improvements. Typical uses include pre-release load validation, failover stress checks and limit verification in cloud setups.

✔

  • Early detection of scale limits and single points of failure.
  • Data-driven basis for capacity planning and cost estimation.
  • Improved resilience through targeted remediation.

✖

  • Results depend on test environment and data representativeness.
  • Complex system states (e.g., caching) are hard to reproduce.
  • Costs and effort for large-scale tests can be significant.

  • Response latency (P95/P99)

    Measure of response times under high load, important for UX and SLAs.

  • Error rate

    Proportion of failed requests under load, indicator of stability.

  • Throughput (requests/sec)

    Maximum achievable throughput before degradation occurs.

E‑commerce shop before Black Friday

Stress test with simulated spike users to validate scaling rules and cache behavior.

API gateway failover validation

Test gateway under backend cluster failures and measure response degradation.

Database cluster size limit

Determine the point at which replication and transaction throughput collapse.

1

Define target metrics and acceptance criteria

2

Create realistic load profiles from production data

3

Set up a monitored test environment

4

Run incremental load tests up to failure point

5

Analyze metrics, logs and traces for root-cause

6

Derive and implement optimization measures

⚠️ Technical debt & bottlenecks

  • Unmeasurable service levels hinder precise evaluation
  • Missing automated test scripts hinder repeatability
  • Monolithic components that cannot be tested in isolation
Database throughputNetwork bandwidthLock and thread contention
  • Mass tests in production windows without coordination
  • Ignoring background processes that occur in production
  • Relying on a single metric signal for assessment
  • Lack of isolation leads to skewed results
  • Incomplete observability prevents diagnosis
  • Skipping cleanup steps after test runs
Knowledge in performance engineering and system metricsExperience with load testing tools and scriptingAbility to analyze distributed system behavior under load
Scalability under loadAvailability and failover behaviorCost of resources and operations
  • • Test environment must sufficiently emulate production behavior.
  • • Budget and time constraints for large-scale tests.
  • • Legal and privacy constraints when using production data.