How to do replication? about siberite HOT 4 OPEN

For horizontal scalability you can add more independent siberite boxes and client will use them in round robin manner.

Replication and fail-over is not supported in siberite. Siberite is a simple service, and for replication it would require adding raft of paxos based solutions. It's nice to have but also will affect the performance.

from siberite.

Simple replication might be achieved by standing up n instances on different boxes/containers/disks and then using a custom client that writes to all of them and reads from all of them with deduplication.

Otherwise, you can store the DB itself on a transparently replicated filesystem?

from siberite.

By the way, there's also this: https://github.com/facebook/mcrouter. In theory if you set that up with Siberite on the backend, you should get replication and sharding with consistent hashing.

from siberite.

Thought I'd add one more project I recently came across from Netflix that does a similar thing but in a somewhat datastore agnostic way: https://youtu.be/57UK46qfBLY.

from siberite.