SQL Server has three types of replication: Snapshot, Merge, and Transaction. Snapshot replication creates a snapshot of the data (point-in-time picture of the data) to deliver to the subscribers. This is a good type to use when the data changes infrequently, there is a small amount of data to replicate, or large changes occur over a small period of time.

Merge replication uses a snapshot to seed the replication. Changes on both sides of the publication are tracked so the subscriber can synchronize with the publisher when connected. A typical use for this type of replication is in a client and server scenario. A server would act as a central repository and multiple clients would independently update their copies of the data until connected. At which time, they would all send up their modifications to the central store.

Transaction replication also begins with a snapshot only this time changes are tracked as transactions (as the name implies). Changes are replicated from publisher to subscriber the same as they occurred on the publisher, in the same order as they occurred, and in near real time. This type of replication is useful when the subscriber needs to know every change that occurred to the data (not point-in-time), when the change volume is high, and when the subscriber needs near real-time access to the changes.