4.2 Fault Tolerance

What you need to know

• Fault tolerance is a system's ability to continue functioning correctly when some components fail.
• The internet is fault tolerant because of redundancy: there are many possible paths between any two points. If a router or cable fails, packets are re-routed through another path.
• Redundant routing means no single point of failure — the internet doesn't depend on any one connection.
• Redundancy has a cost: extra hardware and connections. The trade-off is reliability versus expense.
• The scalability of the internet is tied to this design: adding new networks adds more paths, increasing rather than decreasing resilience.
• Fault tolerance applies beyond networks — backup servers, mirrored data, and duplicated power supplies are the same idea.
• A network is more fault tolerant when the removal of any one node still leaves every other node connected. Exam questions often show a diagram and ask which connection's failure would disconnect part of the network.

Worked example

Picture five routers A–E where A connects to B and C, B connects to D, C connects to D, and D connects to E. If the A–B link fails, traffic from A to E still flows A→C→D→E. The network is fault tolerant for that failure. But D–E is the only link to E: if it fails, E is cut off entirely. D–E is a single point of failure; adding a C–E link would fix it. That's exactly the kind of diagram question the exam uses.

Going deeper

The nuance, edge cases, and connections that turn a 3 into a 5.

• Fault tolerance means a system keeps working correctly — not just stays on — when parts fail. The internet achieves it through redundancy: many paths between any two points, so losing one path doesn't cut the connection.
• The formal idea: a network is fault tolerant with respect to a component if removing that component still leaves every remaining node reachable from every other. The exam's diagram questions ask exactly this: which single link's removal would disconnect the network?
• Single point of failure: a component whose failure alone breaks the system. In a network diagram, a node connected by only one link is a single point of failure for itself. Adding a second link fixes it. The design goal is no single points of failure.
• Redundancy costs money — extra cables, routers, servers. The trade-off is reliability vs. expense. Critical systems (hospitals, financial networks) pay for more redundancy than a home network.
• Redundancy applies to data too: backups, mirrored drives, and data replicated across data centers are fault tolerance for storage. If one copy is lost, another exists.
• The internet's fault tolerance is a consequence of its design history: it was built to survive partial destruction. Packets automatically route around dead nodes because each router only needs to know the next hop, not the whole path.
• Scalability and fault tolerance reinforce each other: adding networks adds paths, which adds redundancy. A bigger internet is a more resilient internet.

Mistakes that cost points

• Confusing redundancy with sending duplicate packets. Redundancy is extra paths (and backup components), not repeated transmissions. TCP re-sends packets that are lost, but that's error recovery, not redundancy.
• Not checking every link in a diagram. The question is usually "which link's failure disconnects the network?" Test each one by mentally removing it.
• Assuming a bigger network is more fragile. The opposite — more nodes means more alternative paths.

Practice questions

Written in the style of the real exam. Try each one before revealing the answer.