Streaming cluster replication

Streaming cluster replication (SCR) takes advantage of the fully-connected nature of clusters and data streaming to produce a low-overhead cluster replicator.

SCR decreases the scheduled replicator overhead for clusters and provides a significant reduction in cluster replicator latency. As changes occur, for example, note changes, unread changes, and folder changes, the changes are captured and immediately queued to other replicas in the same cluster. The result is a more efficient cluster replication.

To use SCR, you must have at least two Domino® servers in a cluster. If your cluster contains any combination of current-release Domino® servers and servers running releases earlier than 8, for the pre-Domino 8 servers, regular cluster replication is used to replicate between current and previous-release servers.

Domino® enables SCR when the following sequence of events occurs:

  • A database is opened
  • Cluster replication replicates a change to another Domino® (current-release) clustered replica
  • All references to the database are closed
  • On future changes in the database that needs to replicate, SCR replicates the change to the other Domino® (current-release) servers.

If SCR encounters any errors during replication, it returns control of the database to the standard cluster replicator to replicate the change and bring the database in sync.

As a change occurs to a clustered server's database, the specific change is "captured" and the change is immediately queued for replication to the source server's Domino® (current-release) cluster mates that have a replica of the database. When the destination cluster-mates receive the change, they apply it to their replicas.

The performance benefits of streaming cluster replication are:

  • SCR does not perform operations, such as database opens and note opens, across the network. Database changes are sent to the replicas with the assumption that the database replicas need the change.
  • SCR does not have to replicate one-to-one with each cluster-mate for a database that has changed.
  • The latency time for replication is immediate with SCR.