On their end, operators were experiencing exponential growth in the traffic of data. To meet these needs, they developed packet mode transmission networks by using ATM or IP technologies. The idea of placing telephony on these new networks in order to obtain economies of scale, and notably at the level of transmissions, came about naturally.

In fact, services evolve much more quickly than transport infrastructures. The transport layer represents the most substantial portion of investments. It is thus important not to cast doubt on the investments in the transport layer when changing a service.

Furthermore, operators want to have a more open architecture that makes it possible to take advantage of the market available.
In view of the complexity of the solutions used in telecommunication networks, it is impossible for one supplier to be the best in all areas. With an open architecture that enhances segmenting the offer, certain industrial companies can develop an in-depth expertise and propose truly innovative offers. Operators thus have the best products available while encouraging competition. Today, in intelligent networks for example, if an operator wants to develop new services, he depends on the platform manufacturer. If the manufacturer is too expensive, the operator cannot make his network progress.

Another advantage of an open architecture: the improvement in the implementation cycles of new services. In the competitive world of telecommunications, one must be able to react quickly. "Open" technologies make it possible to change a component and to implement a new service in a few months (today, the time required is more on the order of one year).

Obviously, new entries do not have the same problems: they do not have to protect their investments. They can thus take more risks and opt for more avant-garde technologies. In the same way, traditional operators are also competing in their market conquest approach at an international level. France Telecom is often the new entry on other markets at the level of its subsidiaries.

The goal is thus to minimise investment costs and to have a better reactivity.

Today, ADSL plays this role over copper wires, or fibre optics in Scandinavian countries. As ADSL is not really profitable for just simple broad band Internet access, it is now a question of providing new services that take advantages of the possibilities offered by this technology.

We first start from an observation: data transmission plays a major role in developed countries (the U.S., Japan, etc.) and this data traffic has required setting up specific transport services using specific ATM or IP technologies.

Four concepts are a deciding factor in choosing technologies (see insert 1). First, it involves using a packet network in the everyday transmission of information (voice, data and video) and of all the services. Next, intelligence must be externalised and distributed. Today, the telephone switch includes all of the communication layers. In NGN architecture, the layers are well identified and linked by interfaces that are clearly defined. It is thus possible to protect the investments in each of the layers, whatever the modifications in the other layers of the network may be. The third concept: a fibre optics transmission network. We already have a core all-optical network but the largest part of the network is still made up of copper pairs. However, applications are hungrier in terms of throughput and in time will require an evolution towards an all-optical network. The last concept: segmenting the market through open interfaces. We define standard interfaces among the different layers. The operators can thus develop heterogeneous networks to take advantage of better equipment, something, which is still very difficult to do today.

Example of a visiophonic service based on the target NGN architecture

Another starting point: access technologies currently co-habit (copper, radio or even satellites) and notably for Internet. Local regulations also have an effect on new technologies, as do short or long-term investment strategies.

In short, all these parameters lead to two different architectures: the "telephony" approach or the "multimedia" approach (see insert 2).

The "telephone network" has been tested by France Telecom with Ericsson and presents characteristics of reliability that are fully mastered by industrial companies.

The "multimedia" or "data" network, which today seems to be preferred by the market, is more an evolution of the IP networks. But quality of service is not guaranteed. Scalability is also one of the weak points of this network: the machines used today have a small capacity and cannot manage a capacity similar to that of the telephone network in a simple manner.

Once again, a solution is not easily found. Two adaptation techniques exist.

First, voice over ATM. It has the support of the majority of transit solution manufacturers because it makes it possible to guarantee a quality of service, as users do not accept any regression. On the other hand, that ATM will be perpetuated at the core of the network is no longer a certainty.

There also exists a second technique: Voice over IP. It works closely with applications and the protocol is a well-established one. Integration with the network takes place easily as soon as IP machines have been set up at the core of the network. Something is still slowing its development: a quality that is not guaranteed.

In fact, as soon as IP will be able to provide a sufficient quality of service, ATM will disappear from the network. This migration will take place in two to three years. Today, experts do think the target is Voice over IP (VoIP).

The NGN on DSL and the multimedia approach is another promising application. The idea is to use the ADSL access already deployed for broad band Internet to offer new services.

For France Telecom, it is a question of offering innovative services to its clients by using ADSL-type broad band accesses. Of course, the idea of introducing mobility in the fixed network is still around. Because the client wants services that we call seamless ones, he does not care whether he has services on a fixed or on a mobile network. His problem is to have his service environment available wherever he may be. This is one of the great challenges for NGN.

NGN Concepts and Legacy Networks

Concept No. 1: using the same packet network for the everyday transmission of information (voice, data and video) and all other services

ATM makes it possible to provide a quality of service, a QoS to which the France Telecom client has become largely accustomed. However, IP, from its end, even if it has become a standard at an application level, does not know how to guarantee quality of service. IPV6 solves the problems of useful addresses and security but the quality of service depends on technologies still in their experimental stages.

The result: the two technologies are currently combined. ATM is used at the transmission level (level 2 of the network layer). ATM is still the basic technology for ADSL. But the network will have to evolve towards IP, which is used in applications.

Concept No. 2: externalising and distributing intelligence

Today, the telephone switch includes all the layers: transmission with the switching matrix, a control component that makes it possible to control connections on this matrix and the service portion that is often included in the control. All of this is found in the same machine. The semaphore network makes it possible to send information, signals and the command information of the service network.

In the NGN architecture, the connection matrix is spread out over the entire packet network. The idea is to isolate the transmission layer by taking out the control, as shown in the diagram. A call server pilots the gateways, which are found at the transmission level. It only ensures the connection from one point to another. These access gateways are the link between the new technologies and what currently exists. They are the interface between the telephone world and the call server that pilots them. For a traditional operator, the problems of inter-operability with existing networks must be taken into account.

Another advantage is that call servers may be grouped in hosting sites as they are not linked to the connection matrix. Operating costs are thus minimised. The operator can also put specialised teams in place on these servers and guarantee an excellent quality of service. He thus solves a human resource problem: today, the people who take care of a switch must have wide ranging skills because technologies used are more and more state of the art.

At the control layer level, what involves establishing a pathway and the application part are also separated. The latter is brought back into the application platforms with standardised interfaces that are called APIs. With these APIs, the market is opened to new companies, which will design the basic components that will allow an operator to build his own applications rapidly.

Furthermore, the network has well-identified interfaces among the different layers available. It is thus possible to protect the investments in each of the layers, regardless of modifications made in the other layers of the network.

Concept No. 3: an optical transmission network

Basic technologies are evolving. For example, we use wavelength multiplexing and we are heading towards an all-optical network core. Today, we are still primarily using the existing infrastructure, or copper, for the entire periphery, which represents 80% of the investment. But we will have to progress, as applications require higher and higher throughputs and the copper network's limits will one day be reached. This migration has in fact started at the level of companies, which already have direct accesses via optical fibres.

Concept No. 4: segmenting the market through open interfaces

We have defined standard interfaces among the various layers. At the level of equipment, this makes it possible to take advantage in each of the areas and to have heterogeneous networks, something which is currently difficult to do.