This change would add the ability to asynchronously replicate data from one cluster to another.
This is useful for fast read/write in a local region that replicates it's data to another region where fast read/writes can also be achieved. The alternate solution (that gives more consistent view of data) is to simply run a cluster that is spread across multiple regions (i.e. 3 regions with RF=3 or 5 regions with RF=5) however the latency implications for certain high volume cases is not desirable.
This change would add the ability to asynchronously replicate data from one cluster to another.
This is useful for fast read/write in a local region that replicates it's data to another region where fast read/writes can also be achieved. The alternate solution (that gives more consistent view of data) is to simply run a cluster that is spread across multiple regions (i.e. 3 regions with RF=3 or 5 regions with RF=5) however the latency implications for certain high volume cases is not desirable.
Design doc: https://docs.google.com/document/d/1hJktjv0VNKnbbIwLVjQ--HNyODwIT8guDv1eyIFZkzI