Course contentsShow
← System Design
Lesson 1190 of 1,91930. Observability: MetricsFree lesson

The RED Method

Measuring Rate, Errors, and Duration for request-driven services to understand service health.

The RED Method

What you'll learn: How to monitor the health of any request-driven service using just three key metrics: Rate, Errors, and Duration.

What is the RED Method?

The RED Method is a focused monitoring approach for services that handle requests. Instead of tracking hundreds of metrics, you concentrate on three fundamental signals that reveal service health at a glance:

  • Rate — How many requests per second is your service handling?
  • Errors — How many of those requests are failing?
  • Duration — How long does each request take to complete?

These three metrics answer the essential questions every service owner needs to know: "Is my service receiving traffic? Is it working correctly? Is it responding quickly?"

Why RED Works

Think of RED like checking a car's dashboard: speed (rate), warning lights (errors), and engine temperature (duration). You don't need to monitor every bolt and wire — just the vital signs that indicate overall health.

Rate uses counter metrics to track throughput trends. A sudden drop might indicate upstream problems; a spike could signal a traffic surge or attack.

Errors measures your service's reliability. Track this as both a counter (total errors) and a rate (errors per second), then calculate error percentage against total requests.

Duration typically uses histograms to capture latency distribution. You'll want percentiles (p50, p95, p99) to understand typical and tail latencies, not just averages.

When to Use RED

RED applies specifically to request-driven services — web servers, APIs, microservices, and RPC handlers. If your system processes discrete requests with clear start and end points, RED gives you immediate visibility into performance and reliability.

Key Takeaway: The RED Method monitors request-driven services through three essential metrics — Rate, Errors, and Duration — providing a complete picture of service health without overwhelming complexity.