In one presence cluster you can have up to 6 servers (1 publisher and 5 subscribers). A presence cluster can be divided into subclusters (Max of 2 servers per subcluster which provide a max of 3 subclusters).
The subclusters can be configured in 3 modes:
In this mode the users are distributed across both nodes in the subcluster. Assuming that we have 3 subclusters with servers names as A1/B1, A2/B2, A3/B3, the distribution algorithm will distribute the users as A1 -> A2 -> A3 -> B1 -> B2 -> B3.
In this mode the users are distributed to the first server in each subcluster with second server as backup. Assuming that we have 3 subclusters with servers names as A1/B1, A2/B2, A3/B3, the distribution algorithm will distribute the users as A1 -> A2 -> A3.
In this mode the users should be assigned to servers and subclusters manually instead of sync agent.
IMP publisher will sync with CUCM publisher to get all IMP enabled users and their information. Also, note that IMP cluster can connect to single CUCM cluster ONLY.
Why do we need subclusters?
They are mainly used to divide the database size into sub databases. This will make the databases more manageable. For example when taking database backup, the backup job might not complete in the specified window for a single large DB. However, with sub-sized DBs, the backup can be completed within time frame specified.