Course contentsShow
← System Design
Lesson 1265 of 1,91932. Monitoring and AlertingFree lesson

RED Method: Rate, Errors, Duration

Exploring request-oriented metrics for monitoring microservices: request volume, error rate, and response time distribution.

RED Method: Rate, Errors, Duration

What you'll learn: How to monitor microservices health through three essential request-oriented metrics: request rate, error rate, and response duration.

The Core Idea

The RED Method is a monitoring framework specifically designed for request-driven services like APIs and microservices. Instead of focusing on infrastructure resources (like CPU or memory), RED zooms in on the experience of requests flowing through your service. It asks three fundamental questions:

  1. Rate — How many requests per second is my service handling?
  2. Errors — How many of those requests are failing?
  3. Duration — How long do requests take to complete?

Think of RED like monitoring a restaurant kitchen: Rate is how many orders come in per hour, Errors is how many dishes get sent back, and Duration is how long customers wait for their food. Together, these tell you if your kitchen (service) is healthy.

Why RED Works for Microservices

Microservices communicate via requests (HTTP, gRPC, message queues). When something breaks, you'll see it in these three metrics:

  • Rate drops → upstream service stopped calling you, or clients are timing out
  • Errors spike → bugs deployed, dependencies failing, or capacity exceeded
  • Duration increases → service degrading, database slow, or resource contention

These metrics are request-scoped, making them perfect for distributed systems where traditional resource metrics don't tell the full story.

Practical Application

For each service endpoint, track:

  • http_requests_total (counter) → derive Rate
  • http_requests_failed_total (counter) → calculate Error percentage
  • http_request_duration_seconds (histogram) → analyze Duration distribution (p50, p95, p99)

Key Takeaway: The RED Method monitors service health from the request perspective—track Rate, Errors, and Duration for every microservice to quickly identify performance degradation and failures in distributed systems.