The system continues to work even when some components or machines fail. Especially important in distributed systems where you can think of a node failing