Shared dependencies make redundancy cosmetic — two systems on one power supply is one point of failure
When two systems appear redundant but share a common power source, network segment, authentication service, or human operator, treat them as a single point of failure with cosmetic duplication rather than genuine redundancy.
Why This Is a Rule
Two servers in the same rack look like redundancy but share a power supply, a network switch, and a physical location. When the power supply fails, both "redundant" servers go down simultaneously. This is cosmetic duplication — the appearance of redundancy without the independence that makes redundancy work.
Genuine redundancy requires independent failure modes: the backup must not fail when the primary fails, which means they cannot share dependencies that could take both down. Common shared dependencies that create false redundancy: power source (same circuit, same UPS), network segment (same switch, same ISP), authentication service (same SSO provider), human operator (same person runs both systems), physical location (same building, same region), data source (same database backing both).
Each shared dependency is a common-mode failure point — a single event that defeats the entire "redundant" design. The number of redundant copies is irrelevant if all copies share a dependency: three servers on one power supply have the same availability as one server on that power supply.
When This Fires
- Designing backup systems, disaster recovery, or high-availability architectures
- Evaluating whether existing "redundancy" is genuine or cosmetic
- After a failure where both primary and backup failed simultaneously
- During any resilience audit of personal or organizational systems
Common Failure Mode
Counting copies as redundancy: "We have two databases, so we're covered." Check: same physical host? Same cloud region? Same maintenance window? Same admin? If any of these are shared, the two databases provide cosmetic duplication, not genuine redundancy.
The Protocol
When evaluating redundancy: (1) List every dependency each redundant system relies on: power, network, auth, human operators, data, physical location, software dependencies. (2) For each dependency, check: do the redundant systems share it? (3) Any shared dependency is a common-mode failure point — a single event that defeats the redundancy. (4) For each shared dependency → either eliminate the sharing (separate power, separate network, separate operators) or accept that the redundancy is partial and the shared dependency is the true single point of failure.