IP technologies were not deployed in large-scale bearer NGNs until 2004 despite having reached maturity five years ago, illustrating that IP bearer networks are not just affected by technology.

NGNs' business services differ greatly from Internet services. NGN voice services, for instance, offer a clear 'yes/no' outcome in that each service is either connected or not and this is characterized respectively by a service charge or a busy tone for which no charge is levied. Conversely, Internet services demonstrate a gray area in terms of service delivery: The aim is that services should be finally received, but they may be interrupted by delays, jitters and packet loss.

When IP technologies are used to bear NGN services, a key concern is that IP networks must adapt to the transition to NGN service business models. It is unsuitable to plan an IP bearer network based on Internet concepts. IP service models are changing and, currently, there are four main challenges to be addressed. The construction of telecom services from Internet services; Operability problems given that services are in "closer coupling" with networks; Expansion difficulties from 3G, NGN to IMS; Strategic partner selection with regard to constructing IP bearer networks.

Challenge in Construction: Tranformation from Internet to Telecom

Previously, many people assumed that IP bearer networks embody a fairly simple concept that connects VPN, DiffServ and TE. In practice, of course, it is not so easy.

Huawei's IP, 3G and NGN project team carries out multi-system networking planning and tests. The team once produced a report with surprising findings that 'NGN can access the IP bearer network in eight modes, each offering different load sharing and fault protection effects'. Further analysis consolidated early test results and it was affirmed that NGN TG uplink interfaces can do the following: adopt the backup or load sharing mode; achieve milliseconds protection switching duration between active and standby TGs, or between active and standby ports of each TG; distribute inter-TG traffic symmetrically in the mesh shape; implement CAC (Connection Admission Control) for NGN voice services.

None of these features can be provided by Internet terminals and services. Beijing, Shanghai and Guangzhou are the three locations in China at which ICPs are concentrated, and most Internet traffic derives from these three nodes. The bandwidth among these nodes is as large as the traffic between class-3 switches can be omitted, and interactive bandwidth is not considered among other nodes.

However, an IP bearer network cannot be planned in this way. The traffic model in such a network adopts a mesh distribution whereby traffic exists between any class-3 nodes. For example, traffic volumes between Chengdu and Guangzhou might surpass volumes running between Beijing and Guangzhou, since there is a significant floating population between Chengdu and Guangzhou that can cause an unusually large amount of traffic. This rule is applicable for MAN IP bearer networks as well and, given this, business models and service models culminate in different requirements for IP networks. Practice shows that technology alone cannot solve all the problems, otherwise NGNs would have been widely commercialized five years ago.

Reliability is of key importance in IP bearer networks, and so accumulated experience is proving to be extremely important in terms of guiding future development. This is especially true given that IP bearing in NGNs is still undergoing initial commercial application, and that NGN/3G networks have yet to be applied and evaluated in large-scale, mature applications.

Challenge in Operation: "Closer Coupling" between Networks and Services

Internet applications do not require real-time resource information in IP networks and therefore they need not be transparent. In contrast, NGN and 3G network-based telecom services must be aware of resource information in the IP bearer network, and this requires that the network be transparent.

Although IP networks are expandable, transparency challenges the operability of the IP network because of topological structure considerations, flow convergence and flow impacts. When telecom networks are borne in IP mode from TDM mode, they should not be completely dependent on the IP network, unlike Internet applications. The QoS of telecom services needs to be guaranteed by planning, management, monitoring and control.

Planning

Telecom service and Interntet application IP bearer networks differ greatly. The former is connection reliant; the latter is not, and targets "openness". This presents design planning problems. The TDM design mode cannot be completely adopted in an IP bearer network as this would obviate flat networking and limit the network's expansion capabilities. Equally, it is unworkable to completely rely on Internet design ideas as the TDM mode's mature telecom service bearing experiences would be rendered unusable.

The logical solution is to draw on and combine both TDM and IP, bearing in mind that the aim of network planning (traffic models and bandwidth) is to establish a definite traffic path in both normal and abnormal conditions. It must be considered that, in the case of any abnormality, the following should occur: Network oscillation should be controlled to a minimum range; Core traffic should not be transferred via convergence nodes; Convergence traffic should not be transferred via access nodes; Local area traffic and traffic spanning large areas should not be transferred via other areas.

Besides service planning, voice traffic model and bandwidth planning is equally important. Worth to note is that Internet traffic is different in character from media streams, signaling streams and PS and multimedia domain traffic, In addition, IP-based signalling may also become an issue in the future. To avoid signaling packet loss due to voice traffic impact, though, careful bearer network planning is required. An IP bearer network oriented to NGNs is far more complicated than the Internet, but many traditional IP providers remain unaware of this.

Management

A major difference between IP bearer and Internet networks is that the former bears telecom services and offers MPLS, DiffServ and FRR, for which a highly competent service management platform is necessary. The alternative is considerably more difficult, i.e. to manually configure and monitor these services by hops via Telnet logon in a network comprising of several hundred pieces of equipment. An IP bearer network should not only consist of an NE management platform, but also one for LSP, VPN and QoS functions. This would locate harmed VPN services, evaluate losses and provide a report consistent with the NGN fault report. It is also essential to ensure that a large-scale IP bearer network can operate.

Monitoring

The monitoring of network resources and operational status would occur during network operation including link and bandwidth resources, flow direction, inter-node time delay, jitters and packet loss rate. When a network index exceeds the threshold, alarms will prompt the user to implement network expansion or optimization. While TG equipment can be used to total outgoing traffic, it cannot do the same for incoming traffic and therefore anticipate resource requirements.

Statistics would also be provided for and be used to monitor traffic from/to: TG-TG, TG-router, TG-server, router-server, and router-router. Together with the other functions, monitoring precision will be optimized - the system would form a massive-scale report and analysis tool that provides detailed data about telecom service requirements.

Control

This specifically refers to the 'control' implemented in terms of excessive IP resource use. ITU-T and TISPAN have recommended the implementation of RACS architecture for this purpose, although it has as yet not been applied commercially. The prevalent current method is to implement CAC via the IP bearer network, TG and softswitch. Softswitch defines the threshold for parallel voice services in the IP bearer network and once the call rate exceeds this, the softswitch will restrict calls and transmit a busy tone. If the QoS of the IP bearer network is seriously compromised, the softswitch will also implement call restriction. Indices for time delay, jitter and packet loss rates that affect the IP bearer network will also be recorded by the softswitch.

Two-level CAC is deployed to meet the basic requirement of voice services, although CAC may not be able to prevent call losses in abnormal conditions. Prevention would require that oscillation be controlled to the minimum degree should network abnormality occur.

Therefore, planning, management, monitoring and control measures show: traffic among TG/AG, servers and routers; Indices for flow directions, time delay, jitter, packet loss and call connection rate. Thus the IP bearer network is operable.

Challenge in Scalability: Development from 3G & NGN to IMS

IP networks adopt the C/S mode in the Internet context and the P2P mode in the NGN and 3G context. In terms of IMS/FMC, they will adopt a P2P and C/S mixed mode, which further separates service bearing from resources control. At present, there are still few service types, and network QoS can be ensured via DiffServ and a large bandwidth.

In the future, there may be millions of different types of services with varying rates and QoS requirements. If this is the case, how can these services be scheduled and controlled?

As a matter of fact, IP networks were first designed to be studied and therefore the requirement for network expandability was not high. It is only in recent years that IP networks have been deployed commercially. Once IP bearer networks are adopted to bear telecom services, expandability needs to be considered. Since an SMS occupying a couple of bytes can bring big profits to a carrier, NGN/3G expandability requirement remains an issue for which there remains no certain answer. IP bearer networks must provide upgradable solutions in order to meet the flexible requirements of future services.

Challenge in Cooperation: Choosing Strategic Partners

IP bearer networks, NGN and 3G form a relatively new service operation models. As service operations develop, IP telecom networks will become more and more closely aligned with borne services that will require much in the way of new IP-oriented equipment.

According to McKinsey's statistics, a new product will lose 33% potential gross profits in its lifecycle if released 6 months behind schedule. Conversely, a new product launched a month ahead of schedule will gain 12% more profit.

In this sense, IP bearer network construction must view that partnerships run deeper than simply choosing a supplier and their equipments. Carriers must consider and select only those suppliers who possess a deep understanding of telecom services, strong innovation capabilities, rapid response and an appropriate TCO strategy.

Arguably, constantly evolving, nascent technologies have actually surpassed people's real needs and their current value is defined by the ability to meet service and customer requirements. Understanding service requirements is an accumulative process, and similarly, technology usability should undergo a dynamic process of experience accumulation and validation.

China's space program can be used to illustrate this point. China has to date launched five spacecrafts. The purpose of doing so was not to test the usability of a given spacecraft since each possessed a single lifecycle. The actual aim was to accumulate experience in the following areas: launch process, space navigation and space crafts as man-carriers.

In brief, the purpose was to strive for constant optimization. This process of accumulation is also applicable to the more down-to-earth but equally cutting edge construction of IP bearer networks.