How much more efficient is it to replicate between 2 local databases on the same CouchDB server instance versus between 2 databases on different CouchDB instances? Does CouchDB still use some form of IPC (e.g. sockets) to accomplish instance-local replication or does it resort to a more efficient method?
Share
Local instance replication (i.e. a replication where source and target are not identified with an URL), happen entirely inside the Erlang VM and are thus a lot more efficient. If either source or target are specified by an url, the replication happens over HTTP.