PQ Under Load: Proving Lx Resilience

Validation methodology 8 min read

New here? ValidKeep governs regulated software with linked atomic files in Git and deterministic npx validkeep verify (no LLM in the audit path). Optional ysbuild helps author specs; verify is the impartial gate either way.

Start with the ecosystem map for the platform stack (assurance profile, landscape profile, CLI, ValidKeep Hub). Then read the atomic model ($D_x$, $R_x$, $C_x$, $L_x$) if notation like $C_x$ is unfamiliar. Operational Qualification (OQ) and Performance Qualification (PQ) are explained in their own articles.

Performance Qualification (PQ) proves implementation resilience ($L_x$) under unpredictable load. Unlike OQ, PQ opens the white box: embedded SDK probes in $CODE$ emit telemetry evaluated by ValidKeep Hub. PQ authors focus on design fragments in $L_x$, not business rule prose in $C_x$. Notation is defined in the atomic model.

White-box module view Module interior shows L sub x logic and CODE with PQ telemetry probes under load. White-box view (PQ) Cx boundary Lx logic / design CODE embedded SDK probes PQ telemetry to ValidKeep Hub Probes prove Lx survives under load
Performance Qualification opens the box: probes in CODE emit telemetry tied to design fragments in Lx.

Why PQ is separate from OQ

Implementation introduces risks black-box tests never see: connection pool exhaustion, lock contention, memory growth under burst traffic, retry storms. OQ proves the contract; PQ proves the engineering survives production-shaped stress. Merging them collapses CSA separation of concerns.

PQTS and probe chain

  • *.pqts.md: strategy (load profiles, soak duration, pass criteria)
  • *.pqtest.md: specs joining probes to $l \in L_x$ fragments
  • pqtscode / pqtestcode: load generators and ValidKeep Hub evaluation scripts

The ysbuild pqtest role is blind to $C_x$ marketing language; it reads $L_x$ for probe placement and technical risk tags.

Per-module decision under fractal framing

Not every module needs heavy PQ. A thin config module may inherit parent release cycles only. A leaf API module with database pooling and concurrent writes earns its own PQ suite. Fractal decomposition lets you qualify where technical risk actually lives instead of one release-wide performance chapter.

Release-bound cycles

On verifyrelease/<Release#>, landscape profile enables release-bound test cycles: ValidKeep Hub listens for formal PQ telemetry aggregated across CRs in the release. Single active verification release keeps evidence coherent for QA sign-off.

CR-101 stress angle

Neonatal dose logic may add database lookups under concurrent ward requests. PQ probes on the calculator service measure p95 latency and error rates while load generators simulate burst order entry. Evidence attaches to $L_x$ fragments describing caching and query strategy, not to OQ boundary cases.


Where to go next