Merge replication involves getting the publisher and all subscribers initialized and then allowing data to be changed at all sites involved in the merge replication at the publisher and at all subscribers. All these changes to the data are subsequently merged at certain intervals so that, again, all copies of the database have identical data.
Occasionally, data conflicts have to be resolved. The publisher does not always win in a conflict resolution. Instead, the winner is determined by whatever criteria you establish.
The primary agents used for merge replication are the Snapshot Agent and Merge Agent:
The Snapshot Agent creates ...