Principle: Fault tolerance
Hybrid approach: fail over to an alternate component and retry
- Requires a failover mechanism, which for communications takes the simple form of changes in the routing database
Consideration: performance
- Unless alternate components are only used as spares, failures decrease performance
Additional algorithmic concepts
- Self-stabilization: ability to get to a good state from an arbitrary state
- Wait-free: every process is able to make progress
- Non-blocking: some process makes progress