RED metrics (rate, errors, duration)

RED metrics are a monitoring method for request-driven services. RED stands for rate, errors and duration. Rate is requests per second. Errors is the number or share of failed requests. Duration is how long requests take, usually tracked as a histogram so you can read percentiles. Together they describe how a service looks to its users.

How it works

Record all three for every service and endpoint, labelled by low-cardinality dimensions such as route, method and status code. For example, an API might show 250 requests per second, a 0.4% error rate and a p99 duration of 320 ms. RED complements the USE method (utilisation, saturation, errors), which describes resources such as CPU, memory and disks rather than requests.

Why it matters in production

A consistent RED baseline means every service has the same dashboard and the same first questions during an incident. RED metrics also map directly to availability and latency SLOs, so alerts can be tied to user impact rather than to host-level noise.