Introduction
CAP is the theorem foundation of distribution system. It is impossible for a web service to provide the three following guarantees : Consistency, Availability and Partition-tolerance.
Description
Consistency
If system response success for a write request, then afterward requests have to read the new data; otherwise, all read requests can not read the new data. Data is strong consistency for caller.
Availability
All request(read/write) must receive response in a while, it can be terminated but not be waiting.
Partition-tolerance
The separated node can serve well in network partition.
Results
If system satisfy AP and the separated node still work will lead to inconsistency, which not satisfy C.
But if system satisfy CP, in order to satisfy C in network partition, the request have to be waiting, thus not satisfy A.
And If system satisfy CA, in order to ready node consistency in a period, the system demand no network partition, so, can not satisfy P.
So, we can not satisfy CAP both in a distribution system(particularly distributed storage system), we can ready two of CAP at least.
In the engineer practice, we will guarantee AP and eventual consistency. The common practice is called asynchronous replication.