Every DataFlow cluster has a central registry of information about the cluster, used to discover member nodes. A DataFlow daemon called Cluster Manager provides this service. It is contacted during graph compilation to make the nodes available, but afterwards it is not involved in the execution of the graph.

A cluster is uniquely identified by its Cluster Manager host name and port. As long as clients have these settings, usually provided by a properties file, they can access the cluster. For more information about how DataFlow determines cluster configuration, see Setting Up Clusters.