-1

My understanding: In essence, A and P are the same thing. Because from the perspective of the entire multi-node cluster, AP is always positively correlated: there is basically no way for us to make design choices such as "sacrificing A to get P" or "sacrificing P to get A".

For example: Can you design a high-availability multi-node cluster that does not allow network partitions (that is: once a network partition appears, it will be unavailable)?

Single-node systems are not considered here, because CAP is a law for distributed clusters.

ASBai
  • 724
  • 2
  • 7
  • 17
  • What are you trying to convey here? How does A and P become the same thing? Partition tolerance is an undeniable constraint while building Distributed system, So how does the example that you pointed out bring any meaning. https://robertgreiner.com/cap-theorem-revisited/ – Haagenti Jan 01 '21 at 12:06
  • So in a distributed system, why is P a feature that cannot be rejected? Isn’t it just because you reject P is equivalent to rejecting A? – ASBai Jan 01 '21 at 16:02

1 Answers1

0

Therefore, the CAP theory can only be interpreted as: when the multi-node cluster has a network partition (P), and it is so serious that no partition in the cluster contains the majority nodes. You can now choose to maintain consistency (C) or maintain availability (A).

ASBai
  • 724
  • 2
  • 7
  • 17