On Jul 11, 2013, at 1:44 AM, Boaz Citrin <[email protected]> wrote: > 1. Performance wise, is it better to run the replication on machine A or B? > (i.e. the source should be the remote machine or the local).
Generally, push replication is more efficient, for a couple of reasons: * It doesn’t leave a socket connection open watching for changes (nor does it poll) * The current replication protocol can push revisions in bulk, but not pull them in bulk, so pull replications end up sending more HTTP requests. > 2. What is the toll on running the tasks on both machines? > (still having unidirectional replication, only from A to B. You mean A runs a push to B, and B runs a pull from A at the same time? I wouldn’t recommend that. At the very least it’ll involve more overhead as both hosts try to write the same revisions to B at roughly the same time, with inevitably a lot of collisions. —Jens
