Course contentsShow
← System Design
Lesson 1320 of 1,91933. Reliability Engineering FoundationsFree lesson

Calculating System Availability from Components

Learn how to compute overall system availability when components are arranged in series (sequential dependencies).

Calculating System Availability from Components

You'll learn how to compute overall system availability when components depend on each other sequentially (in series).

The Series Dependency Pattern

When components are arranged in series, they form a chain where each component must work for the overall system to function. Think of it like a row of dominoes standing upright—if any single domino falls (fails), the entire sequence is disrupted.

In distributed systems, this is extremely common: a client request might need to pass through an API gateway (99.9% available), then hit an application server (99.95% available), then query a database (99.99% available). All three must be up for the request to succeed.

The Multiplication Rule

For components in series, you multiply their individual availability percentages:

System Availability = A₁ × A₂ × A₃ × ... × Aₙ

Real-World Example

Your system has three components in sequence:

  • Load balancer: 99.99% available (0.9999)
  • Application server: 99.9% available (0.999)
  • Database: 99.95% available (0.9995)

Overall availability = 0.9999 × 0.999 × 0.9995 = 0.9984 = 99.84%

Notice how the system's availability is always lower than its weakest link. Even with two highly-available components, that single 99.9% component dragged the entire system down.

Why This Matters

This multiplication effect reveals a harsh truth: adding more dependencies decreases overall reliability. Each additional hop in your architecture is a potential point of failure that compounds the risk. This is why microservices architectures require sophisticated reliability engineering—more services means more multiplication.

Key Takeaway: When components operate in series (sequential dependencies), multiply their availability percentages together; the system can never be more reliable than its least reliable component, and typically becomes significantly less reliable than any individual piece.