Redundancy and Parallel Availability
What You'll Learn
You'll learn how adding redundant components in parallel mathematically increases your system's overall availability beyond any single component.
The Core Idea
When you place redundant components in parallel, the system only fails when all components fail simultaneously. This dramatically improves availability compared to a single component.
The Math Behind Parallel Redundancy
If a single component has availability A, then its unavailability (downtime) is (1 - A).
For N identical components in parallel, the system fails only when all N fail at the same time:
System Unavailability = (1 - A)^N
System Availability = 1 - (1 - A)^N
Real-World Example
Suppose you have a service with 99% availability (A = 0.99):
- One instance: 99% available (3.65 days downtime/year)
- Two instances in parallel: 1 - (0.01)² = 1 - 0.0001 = 99.99% available (52 minutes downtime/year)
- Three instances in parallel: 1 - (0.01)³ = 99.9999% available (31 seconds downtime/year)
Notice the dramatic improvement: just doubling your components jumps you from "two nines" to "four nines."
The Practical Implication
This is why you see:
- Multiple database replicas with automatic failover
- Load balancers distributing across multiple servers
- Multi-AZ deployments in cloud environments
Each redundant component acts as insurance. The more parallel paths to success, the fewer single points of failure.
Important Caveat
This assumes independent failures. If all components share a common dependency (same power source, same data center), they can fail together, nullifying the availability gain.
Key Takeaway: Redundant components in parallel multiply your reliability—system availability equals 1 minus the product of all component unavailabilities, transforming 99% components into 99.99%+ systems.