Performance Testing

Performance Testing

Purpose

Measure and explain performance changes with reproducible evidence.

Use this core loop:

1. Define the decision and metrics.
2. Choose the smallest test that can answer the question.
3. Establish a controlled baseline.
4. Measure the candidate under equivalent conditions.
5. Confirm and classify the difference.
6. Investigate causes before choosing a response.
7. Re-measure and record the conclusion.

The objective is not to force every candidate to look faster. It is to determine what changed, why it changed, whether it matters, and what action the evidence supports.

Non-negotiable rules

• Define the performance question before selecting a tool or editing code.
• Prefer repository-standard benchmark, profiling, and load-test tools.
• Compare equivalent workloads, data, builds, environments, and runtime states.
• Record raw values, units, repetitions, and relevant variability; do not report only percentages.
• Separate observed measurements, interpretations, hypotheses, and confirmed causes.
• Do not automatically replace an approach or roll it back because one result is worse.
• Do not claim causality from a profile, trace, or correlation alone.
• Preserve functional correctness; performance improvement does not excuse incorrect behavior.
• Do not generate load against production or third-party systems without explicit authorization and safe limits.
• Do not claim a benchmark, profile, or comparison was run unless it was actually run.
• Keep optimization changes focused and only make them when the user requested implementation, not measurement-only work.

Workflow

1. Frame the decision

Identify:

• the system, operation, code path, endpoint, or workload under test;
• the decision the measurement should support;