-
Applies to cases like NoSQL.
- In the CAP theorem, it refers to the state of having both A and P.
-
The opposite of ACID Properties in Relational Databases? (Genius naming)
-
Even if a part of the system goes down, the overall system keeps running.
-
The system’s data is constantly changing. Eventual Consistency
-
As long as consistency is eventually guaranteed, it’s okay (loose consistency).
- In cases where strong consistency is required (like ACID Properties), the value read after a write is always updated.
- In Eventual Consistency, this is not the case.
- It’s okay to have a lag between a write and the update of the read value.
-
”Quorum”: a technique used in NoSQL and other distributed systems to maintain consistency of processing.
- It aims to avoid Write-Write conflicts (where one is overwritten) and Read-Write conflicts (where incorrect data is read).
- By simultaneously reading/writing to multiple replicas, it becomes safer.
- Specifically,
- Avoiding Write-Write conflicts: w > n/2
- Avoiding Read-Write conflicts: r > n-2
- (n: total number of replicas, w: number of replicas writing simultaneously, r: number of replicas reading simultaneously)
- There are conditions like these.
- Specifically,