Cisco IP telephony systems are designed for high availability. In order to achieve this, the design must include redundancy for failover and rapid recovery. Cisco Advanced Services offers design considerations and recommendations for high availability at the following URL: http://www.cisco.com/warp/public/437/services/rt59/index.html 
.
The Cisco Unified CallManager plays the key role in maintaining call processing following a failure in an IP telephony environment. This topic describes the following high-availability features that are built into Cisco Unified CallManager:
•
Unified CallManager Redundancy Groups
For additional information on Unified CallManager high-availability features, as well as Unified CallManager failover processing and recovery systems, see Cisco Unified CallManager Failure, Failover, and Recovery .
Unified CallManager Clusters
A cluster comprises a set of Unified CallManager servers (or nodes) that share the same database and resources. Unified CallManager servers can be configured to perform the following functions: database server, TFTP server, or application software server. You can dedicate a particular server to one function or combine several functions on one server, depending on the size of your network and the level of redundancy desired.
Each cluster can have only one database server (also called the first node) and usually one TFTP server (either separate or combined with another function). Cisco Systems recommends that large enterprise networks contain a dedicated Unified CallManager database server with other servers (called subsequent nodes) running the Unified CallManager application software. The Unified CallManager application software performs all call control, including signaling of endpoints, feature invocation, and calling restrictions. Large-scale networks typically use paired redundant application software servers, running in an active-active configuration, with endpoints evenly distributed across the two servers. The TFTP server provides configuration files for the endpoint devices and the associated firmware loads. Large enterprise networks typically use redundant TFTP servers.
Unified CallManager Redundancy Groups
A redundancy group comprises a prioritized list of up to three Unified CallManager servers. You can associate each group with one or more device pools to provide call processing redundancy. Each group must contain a primary Unified CallManager, and it may contain one or two backup Unified CallManager servers. If the primary Unified CallManager fails for any reason, the first backup Unified CallManager in the group takes control of the devices that were registered with the primary Unified CallManager. If you specify a second backup Unified CallManager for the group, it takes control of the devices if both the primary and the first backup Unified CallManager servers fail.
When a failed primary Unified CallManager comes back into service, it takes control of the group again, and the devices in that group automatically reregister with the primary Unified CallManager.
Keepalive Mechanism
A keepalive mechanism is an essential part of an IP telephony solution. Keepalives ensure that endpoints (typically phones and gateways) retain their communications path to a Unified CallManager server. Keepalives not only determine when the primary Unified CallManager server is no longer available, they also determine when the site has become completely isolated from a centralized call control system and must revert to some form of remote survivability capability such as Cisco Unified SRST. Keepalives avoid delays in establishing a call caused by searching for an available Unified CallManager server.