The original four (Google SRE book)

Latency: how long requests take. Traffic: how much demand. Errors: rate of failures. Saturation: how full the service is. Designed for services; not always natural fit for stateful systems or batch.

Advertisement

RED for services

Rate (requests/sec), Errors (errors/sec), Duration (latency). Three signals; remember three things. Captured by any HTTP/gRPC instrumentation. The minimum-viable service dashboard.

Advertisement

USE for resources

Utilization, Saturation, Errors. For CPUs, disks, network. 'CPU at 90% utilization' is U. 'CPU run queue length' is S. 'CPU errors' is E (rare; cosmic rays). Lets you reason about resource-bound systems.