Skip to main content
  1. References/
  2. Architecture Design Basics/
  3. Pattern Taxonomy/
  4. Observability/

RED & USE Methods

·· 149 words· 1 min

🟠 P1 — systematic approaches to choosing what to measure

RED Method (for request-driven services) #

SignalMetricAlert when…
R ateRequests per secondDrops below baseline
E rrorError rate (% of requests)Exceeds threshold (e.g. >1%)
D urationRequest latency (p50, p95, p99)p99 exceeds SLO (e.g. >500ms)

USE Method (for infrastructure resources) #

SignalMetricAlert when…
U tilisation% of resource capacity used>80% sustained
S aturationQueue depth / backlogGrowing steadily
E rrorsHardware/resource errorsAny occurrence

Instinct #

RED for services, USE for infrastructure.

In a design interview, after sketching the architecture:

For each service, I’d instrument RED metrics. For the database and cache layers, USE metrics. Alerting on error rate >1% and p99 latency >SLO.

References #