Notesfoundations🌱 seedling
Redundancy
Also: Hardware Redundancy · RAID · Mirroring
Keep spare copies of a component (a disk, a power supply, a whole machine) so that when one dies, another takes over. It works against independent faults only.
Why it exists
At scale, components fail constantly (see Hardware Fault). Without spares, every component is a Single Point of Failure.
How it works
- Component level: RAID 1 mirrors each write to two disks. Dual PSUs, hot-swappable CPUs, batteries and generators.
- Machine level: multiple servers behind a load balancer, data copied across nodes (see Replication).
- Replace and resync: swap the failed part and copy state back from the survivor.
Trade-offs
| Gains | Costs |
|---|---|
| single failures become non-events | extra hardware; keeping copies in sync is hard |
In practice
A server with mirrored disks keeps running when one disk dies. Reads continue from the mirror while the replacement resyncs.
Pitfalls
- Correlated faults bypass it: the same software bug, the same bad config, or a shared power feed hits every copy (see Software Fault).
- A spare that's never tested may not take over when needed.
Connections
- Part of
- Fault Tolerance
- Builds on this
- Leader-Based ReplicationLoad Balancer
- Mentioned
- ReplicationSoftware Fault
Recall
These cards are also in your review deck, scheduled with spaced repetition.