Optimising next-generation network architecture

Mar 1st, 2003

Next-generation networks need to be highly scalable and responsive to changing service requirements as well as simple, resilient and easy to operate. The most obvious way to accomplish this goal is to focus on how best to utilise recent technology advances such as ultra-long-haul transmission, photonic switching, scalable electrical switching, and distributed control schemes.

It is important to understand how traffic topology and the underlying network topology fit together. Analysing European and North American networks we find the following results: an optical channel will transit 2 to 4 nodes; 50-70% of network traffic is passed through the node; and 94% of nodes have a fibre degree less than or equal to four.

Considering today's predominant network structure (Fig.1-1, below), possible evolution paths (Fig.1-2a & 1-2b), and the optimal solution (Fig.1-3).

Most networks are built on transponder based point-to-point DWDM transport pipes with a reach capability of a few hundred kilometres. At the nodes, carriers typically use standard SDH/SONET technology that allows them to groom and protect the traffic. add/drop multiplexers (ADM) forming protected rings are the predominant choice, while digital cross-connects (DXC) interconnect rings and to groom traffic (Figure 1-1).

While this is a robust and well established architecture, it comes with various inefficiencies. The SDH/SONET boxes are typically connected via short-reach "grey" interfaces to the transponder of the DWDM pipe. These back-to-back interfaces are costly, consume power and floor space and represent an additional source of failure.

Both the SDH/SONET boxes and current-generation transponders do not support true customer transparency or mesh-based shared protection schemes. As global services become more and more important, many end-users want to operate their private network at full line rates over one or several carriers' infrastructure without any protocol or interoperability problems.

Additional cost inefficiencies come from the unnecessary O/E/O conversion of express traffic at each DWDM node inherent with point-to-point systems.

One area where inefficiencies could be improved centres on the electrical switch and interconnect aspect of a network. Connecting point-to-point DWDM links by a next-generation electrical core switch with integrated DWDM interfaces would be a good route.(Fig.1-2a).

This architecture represents a logical evolution of a DXC-based SDH/SONET network and provides several benefits. The integrated DWDM long-reach interfaces eliminate two grey interfaces per wavelength at every node and the switch matrix could be designed to provide some level of transparency.

Fig.2 shows the electrical switch size requirements in a generic pan-European network with 14 nodes. The presence of optical bypass off-loads the electrical switch and provides relief on the scalability requirements of the EXC. In an opaque network, these electrical switches need to scale very quickly to very large capacities, assuming the traffic growth maintains a positive trend.

New-generation ultra-long-haul DWDM systems allow us to set up nearly every point-to-point connection in a terrestrial network without the use of electrical regeneration.

Integrating multi-directional connectivity management in the form of a managed optical patch panel or automated wavelength switches is a logical answer.

Fig.1-2b shows the node that is an integral part of an optical express layer. Transit traffic passes through the node without regeneration, only add/drop traffic is terminated via transponders. Multiplexing and grooming of sub-wavelength traffic is handled by separate electrical cross-connect devices that subtend from the optical express layer.

The main value of this approach is the elimination of unnecessary regeneration in the optical path with significant savings.

But there are also weak points in the stand-alone optical express layer. An express highway that only uses O/E/O conversion as an on/off ramp neglects two important network aspects. First it is likely for nearly all terrestrial networks that there is still some amount of 3R regeneration required due to the fact that one or the other path length exceeds the transparency length of the optical plane. There needs to be a pool of regenerators at intermediate sites.

The majority of existing revenue-generating traffic in the network has sub-wavelength granularity requiring some intermediate grooming for sufficient network utilisation.

Last but not least, it is important to point out that non-integrated electrical grooming switches at the edge of an optical express layer create a boundary between the optical and electrical plane that is not only costly but also unnecessary.

The opaque and the optical express architectures both present significant value propositions to improve networks. A next-generation electrical switch with integrated DWDM optics that provides sub-wavelength grooming becomes more powerful when integrated with a photonic express plane that provides optical bypass. While the optical plane eliminates unnecessary regeneration, the electrical plane provides aggregation, grooming, wavelength regeneration and conversion.

The remaining challenge for the equipment designers is to find the best way to integrate the electrical plane and the optical plane.

Since the next-generation hybrid platforms cover all aspects of high-capacity networking, it is important to give network operators various options for how to migrate smoothly from today's static, inefficient networks towards an integrated architecture. Some operators may find it beneficial to first introduce the next-generation grooming switch. In another cases, the optical express plane with integrated wavelength routing and optical bypass may be the more urgent requirement. Any investment today should only be done with a clear vision of how to migrate towards a fully integrated multi-layer switching and transport infrastructure.

Only such a hybrid approach with integrated optical bypass can avoid all unnecessary regeneration in the network; only a hybrid approach with integrated grooming and selective wavelength conversion can eliminate all back-to-back gray interfaces. And only an integrated hybrid approach can bundle all Layer 1 transport functions.

Figure 3 shows the result of analysis performed on a European national network with 0.5Tbit/s total capacity. In this graph we simply count the number of ports and interfaces required in the different architectures.

In the legacy approach (left) every wavelength is terminated with a transponder. The client side of the transponders are connected with each other via a patch panel for the through traffic or connected to the short-reach interface of a separate electrical switch (blue) for the local add/drop. The amount of short-reach interfaces (grey) represents a substantial part of the network cost.

The opaque approach shown in the middle of the graph makes use of a large electrical grooming switch that sits between the DWDM pipes and comes with compatible integrated DWDM interfaces.

On right side of the graph we see the multi-layer hybrid approach. The introduction of optical bypass reduces the port scalability requirement of the EXC back to the original magnitude and eliminates unnecessary regeneration for all express traffic at the node creating cost savings.

Fig.4 compares the opaque EXC centric scenario with the more advanced hybrid architecture for a network in different build-out stages. Assuming a carrier requires a total network capacity of 0.5Tbit/s we can see that the hybrid approach is more than 30% cheaper than the opaque architecture. This is mainly due to the reduced number of DWDM long-reach interfaces. But more importantly, it can be seen that the cost-over-capacity curve of the hybrid architecture is much flatter and also proves in at half or twice that capacity.

Historically grown from SDH/SONET rings and point-to-point DWDM systems, most networks suffer from many inefficiencies that exist between the different transport and switching aspects of the network. Improvement can be achieved by introducing scalable next-generation grooming switches with integrated DWDM optics. These switches streamline the network and increase network utilisation. They enable carriers to offer new grades of service centred around mesh-based network structures. The integrated DWDM optics eliminate many unnecessary transponders and the associated back-to-back gray interfaces, which brings down network costs.

Another area of improvement is centred in the optical layer of the network. Latest DWDM technology with ultra-long-haul (ULH) capabilities allows operators to eliminate all unnecessary 3R regeneration for express traffic at a node. The optical bypass provides significant cost savings from an equipment perspective and makes the network more future proof to scale.

Maximum value can then be gained by integrating them into a single multi-layer hybrid transport and switching platform.

Stephan Rettenberger
Director, Customer Projects
Marconi Communications Ondata

Rettenberger is responsible for introducing next-gen core switch and transport solutions. Previously he was director of product line management at Siemens ICN.

More in Network Design