Single Point of Failure in Vertical Scaling
What you'll learn: Why relying on one powerful machine introduces critical availability risks that horizontal scaling avoids.
The Core Problem
When you vertically scale—adding more CPU, RAM, or disk to a single server—you're putting all your eggs in one basket. If that one machine fails (hardware failure, power outage, network issue, software crash), your entire system goes down. There's no backup, no redundancy. This is called a single point of failure (SPOF).
A Real-World Analogy
Imagine a restaurant with one super-chef who can cook 100 meals per hour (vertical scaling). If that chef gets sick, the entire restaurant shuts down—no meals served, customers leave unhappy.
Now imagine a restaurant with 10 regular chefs, each cooking 10 meals per hour (horizontal scaling). If one chef gets sick, the other nine keep working. You lose 10% capacity, not 100%.
Why Horizontal Scaling Distributes Risk
With horizontal scaling, you have multiple machines handling requests. If one server fails:
- The other servers continue serving traffic
- You might experience reduced capacity, but not complete outage
- You can add a replacement machine while the system stays online
This fault tolerance is crucial for meeting availability requirements (like "99.9% uptime"). A single powerful server, no matter how expensive or well-maintained, will eventually fail—and when it does, you have zero availability.
The Trade-off
Vertical scaling is simpler initially (one machine = less complexity), but horizontal scaling provides resilience. As systems mature and availability becomes critical, distributing load across multiple machines becomes essential.
Key Takeaway: Vertical scaling creates a single point of failure where one machine's death kills your entire system; horizontal scaling spreads risk across multiple machines, maintaining partial availability even during failures.