In this situation the goal is to minimize the storage capacity requirement for the primary node. A smaller primary database will result in lower server specs, like required RAM. A smaller primary database will also result in faster disaster recovery backups.

To get a smaller database only the most recent messages are kept on the primary node and everything else gets moved to the archive. As a convenience all or some of the messages in the primary store that are copied to the archive node can be stubbed.

This way the capacity requirements will be reduced while still allowing the user to open and search for messages from his primary store.