# RED and USE Methods — Prometheus + Grafana

Source: https://www.skillbyai.com/en/prometheus-grafana/o-methods

> What to monitor.

## Services and resources

Two methods give good coverage. **RED** for request-driven services: Rate, Errors, Duration. **USE** for resources such as CPU, memory, disks and networks: Utilisation, Saturation (queued work), Errors. Google's **four golden signals** (latency, traffic, errors, saturation) combine both views. Build a standard dashboard and alert set from these for every service, so on-call engineers know where to look.

## USE queries with node_exporter

Host resource health.

```promql
# CPU utilisation
1 - avg by (instance) (rate(node_cpu_seconds_total{mode="idle"}[5m]))

# CPU saturation: load per CPU
node_load5 / count by (instance) (node_cpu_seconds_total{mode="idle"})

# memory utilisation
1 - node_memory_MemAvailable_bytes / node_memory_MemTotal_bytes

# network errors
sum by (instance) (rate(node_network_receive_errs_total[5m]))
```

## Standardise across services

Shared metric names and a common dashboard template make every service equally easy to debug.

**Quiz:** What does the S in the USE method stand for?

- [ ] Success
- [ ] Speed
- [x] Saturation
- [ ] Storage

*Answer:* Saturation. Utilisation, Saturation, Errors.