Observability & Evaluation · Data artifact

Profiling Tier Policy

Data artifactObservability & EvaluationObservability & Evaluationarc:ProfilingTierPolicy

A policy defining tiered profiling modes: continuous lightweight monitoring, triggered detailed captures, canary profiling of a traffic fraction, and restricted exhaustive kernel profiling.

Responsibility. Balances profiling overhead against diagnostic value across modes.

Direct neighbourhood (hover for relationship types)

Relationships

configures structural

GPU System Profiler Ch4.4

constrains control

Kernel Profiler Ch4.4

Design guidance

SHOULD run continuous low-overhead profiling 24/7 and trigger detailed captures only for specific diagnoses.
MAY instrument a small percentage of production traffic (canary profiling) for real-world visibility.

Quantitative guidance

As stated by the sources; verify before use.

Canary profiling example: 1% of requests (Ch4.4).

Classification

Patterns: Tiered profilingCanary profiling
Quality attributes: Performance efficiency (ISO/IEC 25010)Maintainability (ISO/IEC 25010)
Risks mitigated: Rejecting all production profiling due to overhead misconceptionProduction slowdown from exhaustive instrumentation

Sources

Ch4.4: T. Nguyen, "Performance Profiling and Optimization," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 4.4. ISBN: 9798244538229.