Site-By-Site Versus Centralized Telephony

Centralization of IT resources has traditionally been the path to lowest TCO, whether we are talking about application processing, or data storage. Centralization provides higher utilization of servers and reduces server costs, reduces operations costs including power, staffing, and management, and provides increased agility to accommodate network changes and new sites. Do these concepts equally apply to IP telephony?

IP telephony partitions telephony system functionality into a number of elements, which themselves can be centralized or distributed as required across a telephony-grade IP network: clients, communications servers, media gateways, and application servers (e.g., for unified messaging and conferencing). Whether centralization or distribution of this functionality is right for you depends on how well either approach meets your business, reliability, functionality, and price/performance requirements.

Factor 1: Remote Site Business Model
If remote offices are employee-centric, economics rather than customer service may be the primary deciding factor for a centralized architecture. On the other hand, if remote offices are run as separate businesses and traditionally look after their own telephony needs (e.g., in some franchise environments), then a distributed model may best fit your environment. In addition, if service functionality and consistency for telephone customers needs to be delivered at your branches under various failure conditions, this requires careful attention to calling scenarios and is easier to achieve with a site-based IP telephony approach.

Factor 2: Business Continuity/Disaster Recovery and WAN Reliability
The key questions are how reliable is your WAN and at what cost, and what is acceptable business impact in case of WAN failures.
There are three general approaches to meeting these requirements. Firstly, you can centralize communications and application servers, with PSTN gateways deployed on a city basis to minimize cost of handling public voice calls. Secondly, you can centralize communications and application servers, but deploy survivable gateways at each site to provide key telephony features and PSTN access in case of network failures. In these cases, the cost benefits of centralizing telephony functionality need to be weighed against the costs of upgrading the IP network to provide the required level of reliability and availability. Thirdly, you can add IP telephony on a site by site basis, with less reliance on the network.

In all cases, you should consider business continuity and disaster recovery requirements for any IP telephony communications server that is handling a large number of users, whether at larger sites or in data centers. Solutions range from cold standby to active-active system designs.

Factor 3: WAN Bandwidth and QoS
Another factor to be considered is the obvious need to provide adequate bandwidth for voice and associated QoS capabilities, remembering that a highly compressed IP telephony call uses 24Kbps. The bandwidth to each site must be able to handle all inter-site data and voice traffic. Under failure conditions, there has to be enough bandwidth to handle priority traffic, including enough CO trunk capacity to handle calls to be offloaded to the phone network. Intra-site calls don’t use up WAN bandwidth, except for a low level of signaling traffic for centralized deployments. Centralization results in increased traffic to access application servers, such as conference bridges and voicemail.

Factor 4: Client Mix
Clients are a major cost of any telephony system. In a new remote office or branch, client choices include wired and wireless telephone sets, and IP telephony and multimedia soft clients. For existing sites, there may be an economically driven desire to retain current digital sets, complementing these with IP telephony clients where required. Digital sets and incidental analog sets (and fax machines) can both be handled by line side media gateways.

Factor 5: Business Economics and Migration Risk
Centralization puts increased demand on the network in terms of bandwidth and reliability, the cost of which needs to be included in any cost-benefit analysis. In addition, building in adequate levels of redundancy in the central site requires additional investments. For some enterprises, it may be more attractive to take a nodal approach to IP telephony, and thus limit the impact of possible failures to a single site. Specifically, in many customer service branch environments, service continuity is paramount and may not be economically achievable with a centralized approach. Finally, leveraging installed base investments in an IP telephony environment can be a very positive contributor to a business case.

Centralize If You Can; Distribute If You Must
In deploying large scale IP telephony systems, you should carefully assess your business needs and the degree of centralization that best meets these needs. Distributed architectures more closely follow today’s approaches and are the lower risk path particularly during the transition period and if you are running a customer-centric network. In many cases, you will opt for a hybrid approach with some degree of centralization (e.g., around the metropolitan area where your head office is located) and distribution (e.g., for off-shore sites where network costs may be considerably higher). Alternatively, you can decide to distribute your IP telephony systems while centralizing your unified messaging and multimedia capabilities. IT

Tony Rybczynski is Director of Strategic Enterprise Technologies at Nortel. He has over 30 years experience in the application of packet network technology. For more information, please visit www.nortel.com.

If you are interested in purchasing reprints of this article (in either print or PDF format), please visit Reprint Management Services online at www.reprintbuyer.com or contact a representative via e-mail at [email protected] or by phone at 800-290-5460.