The amount and nature of any local data that is stored in the replica database affects the resource requirements of the synchronization process. For more scalability, you should partition the data into separate replica databases, where each replica database contains only a subset of the master database data that is necessary for the users that access the replica database. Typically, as the data is partitioned, you achieve more scalability in your overall system. Make sure to consider performance needs when designing the logical and physical data model of the system.