Throughout this chapter we have discussed the importance of redundancy and
load distribution in an AVVID infrastructure, and how this may be achieved using
features supported by Cisco equipment.
We have looked at how several Cisco CallManagers can be clustered together
on a fast network to provide transparent redundancy, and illustrated load balancing
of call processing between each server.
Intra-cluster communication takes place between members of the same
cluster. Information communicated using this method includes the CallManager
Database, replicated using a client server model, and real-time data such as device
registration, using a full-mesh topology. Intra-cluster communication requires a
fast, switched network media to perform.
There is a limitation of eight CallManager servers in a cluster, which must
include a database server,TFTP server, a primary call processing server, and
backup call processing servers (optional), and application servers. If necessary,
these features can be combined onto the same server.The number of devices supported
by each server depends on the processor, the memory of the server, and
the weights of the devices connecting. Regardless of the hardware, no more than
2,500 IP phones can be supported by a single CallManager server, and no more
than 10,000 IP phones can be supported by a single cluster.
To overcome these limitations, it is often necessary to use multiple clusters
within an organization if it is very large or geographically diverse. Inter-cluster
communications take place between separate clusters using the H.323 protocol
over inter-cluster trunks. Unfortunately, many CallManager features are not supported
between clusters.The only features currently supported are basic call
setup, G.711 and G.729 calls, call transfer, call park, call hold, calling line ID, and
multiparty conference.Therefore great care should be taken when designing the
IP telephony network, and unnecessary clusters should be avoided.
The use of Cisco CallManager groups facilitates redundancy, and load balancing
between cluster members can be achieved using groups and device pools.
Three CallManager servers—a primary, secondary, and tertiary—can be part of a
group.A device pool is a group of IP telephony devices sharing the same characteristics,
including the CallManager group associated with it.When a device or
device pool is associated with a CallManager group, it uses the primary server
when available, but can switch to using the secondary or tertiary server in the
event of a failure. Load balancing can be achieved by assigning different device
pools to different CallManager groups.
www.syngress.com
126 Chapter 4 • AVVID Clustering
Voice and videoconferencing devices may also be linked together to improve
scalability and performance through a feature known as cascading.Two or more
Media Control Units can be linked together in order to facilitate larger conferences
and to effectively distribute conference processing and voice/video traffic.
Users only share voice/video streams with the MCU to which they are attached,
and there is only a single voice/video stream between each MCU. Conference
participants can easily invite other MCUs to become part of the conference
through the use of their terminal or by way of a Web browser.
Through the use of features supported by the Cisco AVVID infrastructure,
it is possible to implement robust, high-performance, and scalable multimedia
solutions.