Info
A highly distributed system can provide 2 of: Consistency, Availability and Partition Tolerance at most.
- Consistency: when reading data, every request receives the most recent data or an error is returned
- Availability: when reading data, every request receives a non error response, without the guarantee that it is the most recent data
- Partition Tolerance: when an arbitrary number of network requests between nodes fail, the system continues to operate as expected.
Reasoning
The core of the reasoning is as follows. It is impossible to guarantee that a network partition will not occur (see The Fallacies of Distributed Computing). Therefore in the case of a partition we can either cancel the operation (increasing consistency and decreasing availability) or proceed (increasing availability but decreasing consistency).