Atomic materialization is the default materialization method. It is the easiest method to execute and maintains complete data consistency throughout the materialization process.
During atomic materialization, Replication Server logs in to the primary data server as the user creating the subscription and with the password defined at the replicate Replication Server. Therefore, the user must be defined at both the replicate Replication Server and primary database with the same password. The user also needs the same login name and password as the primary Replication Server.
Logged in to the primary data server, the Replication Server selects the subscription rows using a select with holdlock operation specified by the rs_select_with_lock function. The holdlock performs a repeatable read, preventing other transactions at the primary site from updating the data until the select transaction has completed. The rows are transferred to a materialization queue at the replicate site, where they are applied to the replicate database. You must provide the stable queue with adequate partition space to handle the operation.
Atomic materialization is best for smaller subscriptions where the select with holdlock operation does not last long enough to disturb client applications using the primary database. If the subscription selects a large number of rows, you may choose to use nonatomic or bulk materialization, so that clients at the primary database are not affected.
When data already exists at the replicate database, you can use the no-materialization method.
Atomic materialization allows changes to the primary table but effectively delays data server changes until the activation phase of materialization has completed.
