Infrastructure metrics (CPU, memory, disk, network) tell you the host is healthy. APM tells you the application is healthy.
APM metrics include:
- Request rate — requests per second
- Latency — p50, p90, p95, p99 (percentile distributions, not just averages)
- Error rate — percentage of failed requests
- Apdex score — user satisfaction approximation
- Database query times
- External call latency (downstream services, APIs)
Tools that do APM: Datadog APM, New Relic, Dynatrace, AWS X-Ray, Application Insights (Azure), Jaeger/Tempo (open-source distributed tracing).