A single point of failure (SPOF) is any component whose failure brings down the entire system.
Examples:
- One database server with no replicas
- One load balancer with no backup
- One datacenter in one region
Eliminate SPOFs through redundancy:
- Database replicas
- Multiple load balancers
- Multi-region deployment
In interviews, after drawing your architecture, ask yourself: "If this component dies, what happens?" If the answer is "everything breaks," you have a SPOF.