Content
FLP impossibility principle The theoretical lower limit of consensus algorithm
The paper proposing the theorem was published in 1985 by the three authors of Fischer, Lynch and Patterson. The paper later won the Dijkstra (the one who invented the shortest path algorithm) award.
The FLP principle believes that a purely asynchronous system cannot ensure that the consistency is completed in a limited time when the node is allowed to fail.
Voting example for three people in three rooms
Three people vote in different rooms (the voting result is 0 or 1). Three people can communicate with each other over the phone, but often someone falls asleep from time to time. For example, at a certain time, A votes 0, B votes 1, C receives two votes, and then C falls asleep. A and B will never know the final result within a limited time. If you can vote again, a similar situation happens every time before the result is obtained
Bringing into the computer field means that even in the case of reliable network communication, the lower limit of the consensus problem of a scalable distributed system is unsolvable. That is, the lower limit of reliability is 0%