A CUCM Cluster refers to a group of CUCM servers that share the same version of CUCM, enabling high availability of services, resource sharing, features, and scalability.
In case of a server failure within the cluster, resources are automatically redirected to another server in the same cluster. Each cluster consists of a single publisher and multiple subscribers. Configuration is done on the publisher only, but the Cisco CallManager service must be enabled on every subscriber.
The publisher is also known as the first node, while a subscriber is known as the subsequent node. Only the publisher requires clock synchronization with the lowest stratum NTP server. Subscribers get their time from the publisher through NTP.
CUCM supports two types of clusters – the cluster type can support up to 30,000 IP phones and can be scaled to a maximum of 20 servers. The supercluster or mega cluster requires approval from the Cisco BU and can support up to 60,000 IP phones and be scaled to a maximum of 21 servers.
There are two types of data communication between CUCM servers – database replication and cluster communication. The publisher contains the read/write copy of the database, which is replicated to all subscribers. The database replication is secured using embedded Red Hat Linux and ip tables dynamic firewall.
ICCS signaling is used to replicate run-time data like registration of devices, locations, bandwidth, and shared media resources. This signaling is only between servers that have the “ccm.exe” service, i.e., the call processing agents.
It uses TCP ports 8002-8004. CDR & CAR data are logged by the Call Processing Engine and pushed periodically to the publisher server. The Cisco CAR or third-party billing application server collects data from the publisher.