Applications/Services and Requirements

Table 1 - NGN multimedia services

Within this context, basically three kinds of applications can be distinguished:

·  Elastic applications: Generate discrete media traffic(e.g. file transfers, www)

·  Adaptive applications: Generate continuous media traffic (e.g. video and audio),

· Critical applications: Generate continuous media traffic (e.g., teleteaching, teleconferencing, medical telediagnostics, video/entertainment on demand, and distributed games )

·  intolerant applications (e.g., interactive games, some control applications)

· tolerant applications (e.g., interactive voice applications).

Elastic applications and requirements

In Internet, the bulk of traffic has so far been generated by elastic applications. Because the pieces of data exchanged by such applications are carried over the network as packets that have few time constraints, we call such traffic discrete media. The lack of time constraints on the packets allows the network to view them as loosely coupled. In the Internet, the basic service is provided by considering the data packets independent of each other (such packets are called datagrams). Elastic applications are essentially composed of file transfers.

However, some elastic applications exhibit a degree of interactivity, and therefore have additional performance requirements. For instance, the World Wide Web, the most important application on the Internet today, makes extensive use of on-demand file transfers between servers and clients. Because the perceived "quality" of the transfers depends on the users, their moods, the purpose of the transfers, as well as many other factors, strict performance requirements cannot easily be identified.
 

Figure 1 - The end-to-end delay allocation model for a VoIP PC-phone call.

Table 2 - VoIP round-trip delay allocation and current performance in milliseconds.

Adaptive applications and requirements

In addition to discrete media, applications increasingly make use of continuous media, due to advances in coding technology and the availability of multimedia computers. In continuous media, specifically video and audio, the data have intrinsic temporal and spatial relationships that must be respected for these forms of data to make sense. The performance requirements of continuous media are closely linked to their perceived quality.

Techniques have been devised whereby the playout quality of continuous media is adjusted to match the instantaneous capabilities of a system, and in particular of a network. Such techniques, used by adaptive applications, allow the use of multimedia applications on best-effort IP networks-Internet. However, even the best adaptive techniques are powerless when facing the poorest conditions in a network; and, as a consequence, guarantees cannot generally be given as to the quality delivered by adaptive applications on best-effort networks.
 

Figure 2 - Voice quality as a function of packet loss rate

Critical applications and requirements

On the other hand, in order that distributed multimedia applications become ubiquitous, especially in a commercial environment, there is a need for a communication platform that is able to provide better control and guarantees over performance. Especially for commercial use of applications such as

·     teleteaching,

·     teleconferencing,

·      medical telediagnostics,

·     video/entertainment on demand, and

·     distributed games,

For these applications, starting a communication session may be worthwhile only if some minimum performance can be guaranteed throughout its duration. These can collectively be called critical applications. Critical applications can further be classified into intolerant applications, which do not tolerate any deviation from their expressed requirements (e.g., interactive games, some control applications), and tolerant applications, which essentially have nominal requirements but use adaptive techniques to deal with occasional violations of these requirements (e.g., interactive voice applications).

Typically, the requirements of critical applications can be expressed as a (sub)set of values representing

·       bandwidth,

·       delay,

·       jitter, and

·       packet loss rate

constraints for the network. For the VoIP application, the influence of some of these QoS parameters is in Table 2 and Figure 2. In order to be able to meet these constraints throughout the lifetime of a communication session, the components of the network (or at least the subset representing those "manipulating" the packets of the session) must be aware of their values and must cooperate in taking actions to enforce these bounds. To that end, in order to enforce the traffic constraints of a session, a network component has to take part in some or all of the following general activities:

• admission control,
• resource reservation,
• packet scheduling,
• traffic policing, and
• signaling.

It should be clear that because of their very different characteristics, elastic, adaptive, and critical applications cover separate regions of a wide spectrum of demand. In other words, these different types of application complement, rather than compete with, each other. A telecommunications infrastructure will therefore have to support all of them simultaneously and, if possible, seamlessly. This calls for the design of a network where the overhead associated with each type of application is no greater than necessary. In the next section we review potentially promising approaches to providing such support in the Internet. The common perceived conjecture is that these, in combination, may provide the sort of support required by the spectrum of applications.