IP Telephony Network Deployment Overview

Before covering Cisco IPC Express-specific deployment details, you should understand a little more about general IP telephony networks and some of the trade-offs when selecting what type of network fits your business best. This fundamental understanding is needed because Cisco IPC Express networks follow the same general architectures and trade-offs. They also represent a subset of the larger canvas of IP telephony options you may choose from for your business or branch office. Another reason to understand general IP telephony networks is because you can mix and match systems of various types in the same network. If you are already familiar with IP telephony network architectures in general, you can safely skip this introductory section, and proceed directly to the Cisco IPC Express deployment models covered in the next section.

The brain of a telephony network is the call control (or call processing) component. This component of the network generally can be located anywhere in the network in one or multiple places. It provides call features to phones such as dial tone, digit interpretation to implement a dialing plan, and setting up and tearing down voice calls (or, more technically, a speech path or media stream) from the calling user to the called user. The call control component also manages supplementary features such as call hold, transfer, conference, music on hold, call waiting tone, and the myriad voice call features you are already familiar with.

When considering where and how call processing is provided to users and phones, IP telephony networks can be classified broadly into the following types. Each of these is discussed in the subsequent sections:

• Single-site or standalone network
• Centralized network
• Distributed network
• Hybrid network

Single-Site or Standalone Network

Single-site or standalone networks are businesses or networks where all the employees are located at a single site, as shown in Figure 2-1. By definition, these types of offices or networks are predominantly small, often fewer than 30 people, and occasionally perhaps up to 100 or 150 people. Any organization larger than this most likely has multiple geographic sites.

Figure 2-1. Single-Site Call Processing Network

In a standalone deployment, a single instance of the call processing component is resident in the equipment at the site. In rare cases, two instances of the call processing component could be implemented in a standalone site purely for redundancy reasons. However, this is seldom required or cost-effective for such a small site. Having one or two phones directly connected to the Public Switched Telephone Network (PSTN) (or connected via a power failover port on the router, which you learn about in Chapter 6, "Cisco CME PSTN Connectivity Options") usually offers a backup mechanism if your business is of such a nature that redundancy is an absolute requirement.

Centralized Network

In a business with multiple geographic sites, it is very often the case that one site is larger, or more central, to a number of smaller, or remote, sites. In a centralized network topology, shown in Figure 2-2, the call processing component is located at this central or larger site, often the headquarters or main location of the business. This component provides service to all employee phones at all sites, using the network that exists between them.

Figure 2-2. Centralized Call Processing Network

In a centralized deployment, a single instance of the call processing component is resident at one of the sites, and all the other sites connect to this component across the IP network between the sites. Calls are made via IP messaging between the remote devices (IP phones and PSTN trunks) and the central call processing component.

Centralized networks may include multiple instances of the call processing component for redundancy reasons or, in even larger networks, for sheer call capacity and load balancing. In the centralized deployment model, these call processing components are always resident (co-located) at the central site.

Distributed Network

In a network of multiple geographic sites, it is possible that each site is of roughly equal size or importance and that the site-specific network topology is duplicated at each site, as shown in Figure 2-3. In this kind of network, each site has a call processing server that provides services to the community of employees (IP phones) co-located at that site.

Figure 2-3. Distributed Call Processing Network

A distributed deployment, therefore, has multiple instances of the call processing componentone at each siteand each resides at its own site. Intersite calls are completed via peer-to-peer IP messaging between the two call processing components involved in the call.

This kind of network architecture has implicit redundancy, because each site has a call processing component. A network or server outage at Site A does not affect calls made at Site B. The distributed model also has implicit call capacity scalability and load balancing, because each site added to the network has its own call processing capacity. These are some of the strengths of the distributed network architecture.

Hybrid Network

The practical realities of most multisite networks often preclude a design that is either perfectly centralized or completely distributed. Instead, many networks are a hybrid of the two designs. A large network of multiple sites typically contains a number of larger locations where there are enough employees to warrant dedicated on-campus application servers (including a call processing component). A large network also sometimes has a vast number of remote sites that are much too small for this investment in equipment and management.

Reliability and availability constraints also make a purely centralized network less desirable because the single call processing component represents a single point of failure for the entire network. A small number of duplicated call processing servers distributed among a handful of key sites (each serving a larger number of remote sites) provides the best overall network availability service and coverage for large networks. A small number of servers is also cost-effective to deploy and manage.

Duplicating call processing components at multiple (but not all) sites results in most multisite networks being designed with aspects of the centralized and distributed models. This is especially true for larger networks with more than 10 to 20 sites. In this case, the attractiveness of centralized management leans toward the centralized model, while at the same time capacity and redundancy considerations necessitate a partly distributed model. This type of hybrid network includes the following characteristics from each individual call processing model:

• Centralized A number of remote sites draw call processing services from another hub site across the network. The number of remote sites in large networks may be substantialin the range of 100 to 500 sites per centralized hub.
• Distributed A number of sites have call processing servers (and each of these often has a redundant configuration, or cluster). They act as hubs to the rest of the network using a centralized design. Calls between the hubs, though, represent peer-to-peer (or intercluster) calls between distributed call processing components. The number of hub sites is usually relatively small, perhaps 5 to 20, even for fairly large networks of more than 1000 sites.