Management and Governance
Amazon CloudWatch
Amazon CloudWatch collects metrics, logs, traces, dashboards, alarms, and operational signals for AWS workloads. A GenAI platform should correlate model, retrieval, agent, safety, cost, and user-experience measures here or through integrated observability tooling.
Key points
- Custom metrics represent application quality and business signals
- Metric alarms can invoke automated actions
- Logs Insights queries structured log fields across selected groups
When to use it
- Alert on throttling, p95 latency, and guardrail intervention changes
- Dashboard token cost alongside retrieval quality and completion rate
Exam tips
- Choose CloudWatch for operational telemetry and CloudTrail for audit history
- Protect and sample prompts in logs because observability can become a sensitive-data leak