In data groups that are configured for user journal replication, default values use positional replication. In positional file replication, data on the target system is identified by position, or relative record number (RRN), in the file member. If data exists in a file on the source system, an exact copy must exist in the same position in a file on the target system. When the file on the source system is updated, MIMIX finds the data in the exact location on the target system and updates that data with the changes.
User journal replication processes support the update of files by key, allowing replication to be based on key values within the data instead of by the position of the data within the file. Key replication support is subject to the requirements and restrictions described below.
Positional file replication provides the best performance. Keyed file replication offers a greater level of flexibility, but you may notice greater CPU usage when MIMIX must search each file for the specified key. You also need to be aware that data “collisions” can occur when an attempt is made to simultaneously update the same data from two different sources.
Positional replication is recommended for most high availability requirements. Keyed replication is best used for more flexible scenarios, such as file sharing, file routing, or file combining.