Redundancy and Fault Tolerance

Part of: Computer Systems & Networks

Keeping Working When Things Break A system is fault-tolerant if it keeps working even when some parts fail. The internet is built to survive failures of cables, routers, and even whole data centers. The key technique is redundancy : providing more than one way to reach a destination. Redundant Paths If there is only a single path between two points, then one broken link cuts them off. When the network has multiple independent paths , a failure on one path can be bypassed by rerouting packets along another. The more redundant the connections, the harder it is to disconnect any two points. Testing Tolerance by Removing a Node A simple way to reason about fault tolerance is to ask: if this router fails, can the source still reach the destination? We remove the failed node from the graph and check whether a path still exists. Redundancy Is a Trade-Off - More paths mean higher reliability but also more cables and equipment to maintain. - Redundancy applies beyond links: servers, copies of data, and power supplies are often duplicated. - A network with no redundancy has single points of failure , where one outage breaks everything. The Big Picture Fault tolerance is not about preventing

Challenge: Survive a Router Failure