The widespread adoption of Internet protocol (IP) technology has irreversibly changed the networking landscape. Distribution of television content can now be done in a common, IP-centric infrastructure denoted as "triple play," where data, voice, and video share the same underlying infrastructure.

The key success factors for distribution of TV services are as follows:

• Guaranteeing that only the paying end users receive a particular channel
• Assigning high priority to television services
• Taking a cost-efficient approach to the configuration of access nodes
• Establishing a clear ownership of the customer-premises equipment (CPE)

It is important to notice the distinction between Internet connectivity and IP-based networking. Internet connectivity, IP telephony, and IP television (IPTV) distribution are all separate services. The only thing they have in common is that they all use the same underlying IP network.

Properly designed, TV distribution utilizing IP technology can be easily separated from an end user's Internet access. Hence, licensing issues and measures to restrict uncontrolled re-distribution of content over the Internet do not need to become as major a concern as they are in situations in which all traffic is handled as a single service.

For the network owner, flexible configuration of end-user devices on a mass scale is a critical issue, along with the possibility of hosting a number of service providers in the networks.

The amount of headroom for expansion is another important topic in terms of the number of services, the capacity of each service, priority issues, and the total number of users that can be handled by one network operations center (NOC).

The answer to these requirements in the context of the current technology has been rather unclear. But with correct network design, containing priority mechanisms, and multicast functionality, an IP networking infrastructure is today the most attractive alternative for true triple-play networking. The most demanding challenge has undoubtedly been hosting multichannel television distribution.

Television is a highly demanding network service. Transmission control protocol/IP (TCP/IP) was not designed to cater to such time-critical and bandwidth-hungry bit flows. The prime concern here was to correctly transfer each and every bit of information.

Extra delay is, of course, unacceptable when dealing with real-time traffic such as telephony and television. Real-time traffic protocols for the Internet have therefore been developed.

One important step has also been the introduction of priority mechanisms. The dominating technology here is called "DiffServ," giving time-critical traffic a priority labeling when it enters the network.

Telephony (voice) traffic is commonly labeled with the highest priority in the network, followed by video and audio services.

Other factors such as latency (the total, accumulated network delay) and jitter (a variable packet reception rate) also affect quality. These quality issues are generally solved by adding large-capacity headroom in the network (additional data bandwidth). But it is impossible to implement more than a few additional TV channels in the network solely by adding more headroom. We may add here that this is a question of not only technology, but also of operational economy.

The solution to the accumulated capacity problem is a procedure known as "multicasting," in which bandwidth-hungry traffic such as TV channels are sent once and directed only to the receiving party specifically requesting them, thereby minimizing unnecessary distribution. Most traditional traffic in IP networks is of unicast type, where every user requests his or her own bit stream from the source.

Multicast was intended to revolutionize the Internet but failed to do so, mainly because multicast traffic did not fit the volume-based business model of most IP network operators. In a private network such as a triple-play access infrastructure, the situation is different. Here, traffic optimization is important to keeping capacity development within reach. Multicasting also implicitly brings forward the need to control traffic at the network layer (layer 3 [L3]) in the access segment of the network—but it will not lead to the same complexity in the network terminals (i.e., set-top boxes) installed at each end user as a link layer (layer 2 [L2]) approach would do.

L3 is the recommended architecture when designing a triple-play network to ensure flexibility and allow several service providers to share the same network. A major advantage is the increased level of control offered by an L3 architecture when combined with multicasting.

The most critical question for any access network operator today is undoubtedly whether service configuration should reside in the end-user device (CPE). This will affect not only which individual services will have the potential to become profitable, but also the associated price floor for new services (which is defined by the cost for setting up and maintaining the service). Furthermore, the decision will influence the level of control over services that is possible, as well as security and reliability issues.

It is not wise to leave these critical devices in the hands of the end user. CPE units can be dropped on the floor or disassembled and may be costly to replace.

Another important question is how much should be hard-wired in the CPE. One of the worst cases for IP telephony may prove to be VoIP hard-wired or awkwardly implemented directly into the CPE.

The best solution is to establish a clear boundary between the responsibility of the network owner and service provider, and to do it at the outgoing ports of the CPE. The complexity of the hardware is reduced and the network investment will be secured for any future changes in the service offering.