Re:Errors in Database Systems, Eventual Consistency, and the...
by kennu on Wednesday April 07, 2010 @05:58PM
(#1248552)
I would interpret that as a compromise where C, A and P are divided into some ratio (you still don't get 100% of each). I understand Cassandra does that well if the application fits the model.
The point of database partitioning is that a transaction that doesn't span partitions only touches a constant number of servers (usually 2 or 3).