Course contentsShow
← System Design
Lesson 1263 of 1,91932. Monitoring and AlertingFree lesson

Four Golden Signals: Latency, Traffic, Errors, Saturation

Learning Google's core metrics framework for measuring system health across request performance, volume, failure rate, and resource usage.

Four Golden Signals: Latency, Traffic, Errors, Saturation

You'll learn Google's foundational framework for measuring system health through four essential metrics.

The Core Idea

The Four Golden Signals, popularized by Google's Site Reliability Engineering book, provide a simple yet comprehensive framework for monitoring any user-facing system. These four metrics give you a complete picture of system health:

  1. Latency – How long requests take to complete
  2. Traffic – How much demand your system is handling
  3. Errors – The rate of failing requests
  4. Saturation – How "full" your service is

Think of it like monitoring a restaurant: latency is how long customers wait for food, traffic is how many customers arrive, errors are orders that come out wrong, and saturation is how close you are to running out of tables or kitchen capacity.

Why These Four?

These signals cover the essential questions about any system:

  • Is it slow? (Latency)
  • Is it handling load? (Traffic)
  • Is it breaking? (Errors)
  • Is it about to break? (Saturation)

Together, they provide early warning signs of problems. High latency might indicate saturation. Rising errors could correlate with traffic spikes. Monitoring all four helps you understand both what's happening and why.

Practical Application

For a web service, you'd track:

  • Latency: p50, p95, p99 response times
  • Traffic: requests per second
  • Errors: 5xx error rate, timeout rate
  • Saturation: CPU utilization, memory usage, queue depth

These metrics work at every level—from individual services to entire distributed systems—making them universally applicable whether you're running three microservices or three hundred.

Key Takeaway: The Four Golden Signals (Latency, Traffic, Errors, Saturation) provide a complete, focused framework for system health monitoring that applies to virtually any user-facing service.