Fault Tolerance
What is Fault Tolerance?
Fault Tolerance = The ability of a system to keep working even when some parts fail.
In other words: 👉 Your system does NOT go down just because 1 server, 1 database, or 1 component crashes.
Why is Fault Tolerance important?
Because hardware fails, networks fail, software crashes, power cuts happen — but your users expect the app/website to work anyway.
Examples:
- Your web server dies → Users still expect to access the website
- Your database crashes → App still needs to read/write data
How do we improve Fault Tolerance?
Let’s break it per component clearly:
| Component | Problem if it fails? | Solution to tolerate the failure? |
|---|---|---|
| Web server | Users can’t access website | Deploy multiple web servers behind a Load Balancer. If 1 dies, LB sends traffic to the others automatically. |
| Database | App can’t read/write data | Use Primary-Replica setup (Master-Slave). If primary DB fails, failover to replica DB (some DBs like PostgreSQL, MongoDB have auto-failover features). |
| Application server (backend logic) | Business logic is unavailable | Design app servers to be stateless → You can start new app servers easily and balance traffic. |
| Region / Data center | Entire location goes down (earthquake, power cut) | Deploy in multiple regions or availability zones. E.g., AWS offers Multi-AZ deployment → if 1 zone fails, traffic shifts to another. |
| Cache server (like Redis) | Cache unavailable | Use replicated caches or configure cache warm-up to restore quickly. |
Core Tools / Concepts that improve Fault Tolerance
- Load Balancer — Distributes traffic to healthy servers
- Auto-scaling — Automatically adds servers if load increases
- Database replication — Copies of database (read replicas, failover replicas)
- Health checks — Regularly ping services to see if they are alive
- Failover mechanism — Automatic switch to backup when primary fails
- Stateless servers — Any server can handle any request (no session stickiness)
- Multi-region deployment — Your app runs in multiple geographic locations
A simple example
You have:
- 1 Load Balancer
- 3 Web Servers (behind LB)
- 1 Primary DB + 1 Replica DB
Scenario:
- Web Server 1 crashes → LB routes users to Web Server 2 and 3 → No downtime
- Primary DB crashes → App switches to Replica DB → No downtime
= Fault tolerant system.
Golden sentence (for interviews)
“Fault tolerance is about ensuring redundancy and automatic recovery, so the system remains operational despite component failures.”