Three typical CallManager deployment scenarios exist:
Each scenario has different ways of using gateways and gatekeepers.
Single Site Deployment
In the single site model, a central call agent serves one location, such as a building or campus. In small sites, this might be a Cisco CME. Larger sites might use a CallManager cluster. All telephones calls outside the site are sent through the PSTN. Telephone calls within the site traverse the LAN or perhaps metropolitan-area network (MAN). Figure 1-4 shows a typical single site deployment.
Figure 1-4. Single Site Deployment
The example shows a small CallManager cluster used as the call agent. In this deployment, gatekeepers are not needed. Gateways connect the LAN to the PSTN, and sometimes to a PBX. For added redundancy, two gateway routers are used. The G.711 codec is used, because bandwidth is not an issue. The gateway Digital Signal Processor (DSP) resources provide conferencing and media termination point (MTP) resources. To simplify gateway configuration, MGCP is used as the call control protocol. If some specific H.323 or SIP functionality is needed, you could use those protocols instead.
Multisite with Centralized Call Control
A company uses this type of deployment when it wants to include multiple sites in its VoIP network, but centralize the call agents. Typically, a CallManager cluster is located at the headquarters or data center site, with perhaps redundant servers at a backup site. The IP phones in remote locations register to the centralized CallManagers, in a hub-and-spoke type arrangement. Remote sites are connected in a full mesh across an IP WAN. Both call signaling and IP voice are sent over the WAN. PSTN access is usually still provided locally. Figure 1-5 illustrates this type of scenario.
Figure 1-5. Multisite Deployment with Centralized Call Control
The gateways at each location interface with the PSTN, the LAN, and the IP WAN. They might also interface with a PBX at some sites. They provide backup call processing by using SRST if the CallManagers are not available because of WAN failure. You can use the G.729 codec across the WAN; use quality of service (QoS) for best results. A gatekeeper is not required for call admission control, but you can use it for dial plan resolution if many remote sites exist.
Multisite Deployment with Distributed Call Control
If the number of remotes sites grows unwieldy, or sites are widely distributed, you can use a multisite scenario with distributed call control. Each "site" in this scenario might be one of three types:
Figure 1-6 shows an example of this type of network.
Figure 1-6. Multisite Deployment with Distributed Call Control
Multiple CallManager clusters are used, and gatekeepers are used for at least call admission control and address resolution. You can add a directory gatekeeper to scale the gatekeeper implementation. If gatekeepers are used, H.323 gateways are used also. Sites are connected across an IP WAN, with PSTN backup in case of WAN failure or insufficient bandwidth. You typically use the G.729 codec across the IP WAN. Use QoS for best results.
Part I: Voice Gateways and Gatekeepers
Gateways and Gatekeepers
Part II: Gateways
Media Gateway Control Protocol
Session Initiation Protocol
Connecting to the PSTN
Connecting to PBXs
Connecting to an IP WAN
Influencing Path Selection
Configuring Class of Restrictions
SRST and MGCP Gateway Fallback
Using Tcl Scripts and VoiceXML
Part III: Gatekeepers
Part IV: IP-to-IP Gateways
Cisco Multiservice IP-to-IP Gateway
Appendix A. Answers to Chapter-Ending Review Questions