Multihaul DWDM networks emerge

Th 211920

By Roy Bowcutt and Sorin Tibuleac

A combination of service provider organizational structure and existing technology has driven the architectures of most existing optical networks. Service providers adapted their organizations to fit the network segmentations of access, regional, and long-haul and defined the lines of demarcation between the network segments. In this context, a particular technology was considered to be well suited for a specific network segment.

As a natural result, equipment providers developed hardware targeted at these individual network segments using existing technology to satisfy carrier expectations. Suppliers also created EMS/NMS platforms that managed their segment of the network (and only their segment of the network), which reenforced the lines of demarcation.

However, the requirements placed on modern networks have blurred the lines between the traditional access, metro, regional, and long-haul network segments-as well as the equipment attributes necessary to meet these requirements. Rather than operating distinct, interconnected network segments with nearly unique requirements, carriers now find themselves operating what can be viewed as a single, multihaul network in which, for example, technologies originally considered “long-haul” may best meet “metro” requirements in some parts of the network-but perhaps not all.

This evolution in network operation and a need to keep costs down has sparked demand for similarly evolved network elements-multihaul WDM systems that enable carriers to flexibly deploy the right combination of technologies to meet customer and functional demands in different parts of the network.

The traditional network segments and the technologies typically encountered within each of them are well known.

The access network required low wavelength counts, short distances, and a wide variety of client interfaces. As a result, CWDM fit well into this model with its low cost, limited transport capability, and no need for network amplification. This segment, however, required a variety of client interfaces, including OC-3/12/48, Gigabit Ethernet, 10/100 Ethernet, Fibre Channel, ESCON, and even some copper interfaces (DS1, DS3, etc.).

As the most recently defined network segment, the metro segment required transmission distances of 200 to 400 km. Occasionally, the metro network was required to pick up customer traffic directly (bypassing the access network), resulting in a network segment with special needs. Initially, transmission technologies such as directly modulated lasers (DMLs) and single-pump EDFAs fit well into the metro network. As distance demands grew, externally modulated lasers (EMLs) and dual-pump EDFAs were introduced. The metro network typically required the same optical client interfaces available on the access network (OC-3/12/48, Gigabit Ethernet, etc.), plus the ability to aggregate 2.5-Gbit/sec traffic to 10-Gbit/sec data rates.

Before the introduction of the metro network, the regional networks were required to pick up aggregated services from the access network and distribute the services to a long-haul network. The regional network required transmission technologies suited for distances up to 1,000 km, such as EMLs and Mach-Zehnder interferometer transmitters, with data rates focused at 10 and 2.5 Gbits/sec. The network required amplification and was well served by dual-stage EDFAs with mid-stage access for dispersion compensation. Occasionally, Raman booster amplifiers were used to enable longer spans. Services were typically added using a client interface provided directly from the metro or access segments.

The long-haul networks demanded higher performance, and providers were willing to pay for it. As a result, technologies such as lithium niobate modulated lasers and hybrid Raman/EDFA amplifiers fit well into this network segment. Since these networks were expected to transmit large amounts of data over long distances, there was a focus on only a few client interfaces. Managing dispersion compensation was very important, but the limited number of add/drop sites in a long-haul network enabled flexibility in mapping dispersion tolerances.

Many, if not most, modern networks no longer fit these stereotypes. Service providers now need to mix and match features typically associated with one or the other of the previously explained network applications categories. This point is illustrated by a few examples:

1. A core metro ring in an incumbent local exchange carrier or a major cable multiple systems operator (MSO) network in the northeast United States may have short distances between add/drop locations, typical of metro networks, but a large number of add/drop nodes, which requires optical performance typical of long-haul systems.

2. An independent operating carrier providing voice, video, and data services to rural communities in the western U.S. will span distances typical of long-haul networks but have capacity requirements similar to metro core networks in a large city.

3. An ISP may require large transmission capacities and possibly 40-Gbit/sec data rates to interconnect core routers. Such performance is typically available from high-end long-haul systems, but total optical distances fit in the metro or regional categories.Th 211920

Figure 1. The modern multihaul WDM network requires mixing and matching technologies throughout the network-and often on the same ring.

4. The geographical expanses of different countries, the distribution of cities within each country, and their capacity requirements lead to differences in network classifications among carriers. A long-haul network in one country may look like a regional network, or even a metro network, in another country.

Drilling one level deeper into each of these examples, one usually finds that even within the same DWDM ring there are diverse requirements in terms of span length, total link distance, transport capacity, and add/drop capacity. While one network node may be a major hub that requires high add/drop capacity and optical switching capability to and from other rings, another network node on the same ring may operate at a low add/drop capacity. A SONET ring may require frequent termination of the associated wavelength along a DWDM ring; thus, a short-distance transponder is adequate for this service. The same DWDM ring may also carry video services operating on wavelengths that traverse the entire length of the ring. These wavelengths are better served by longer-reach transponders to avoid expensive optical/electrical/optical (OEO) regeneration.

The diversity of span lengths and losses can be supported with different amplifier types deployed on the same ring. Shorter spans may use single-stage EDFAs, while longer spans may require dual-stage EDFAs with mid-stage access for dispersion-compensation modules (DCMs). Extended span lengths may use a Raman amplifier in addition to an EDFA.

Consequently, for practical DWDM network design, even within a single ring there is a need for multiple types of transponders, add/drop modules, and amplifiers.

The challenges just outlined inevitably lead to the concept of a multihaul network. Multihaul combines access, metro, regional, and long-haul into a single optical network (see Figure 1).

The typical multihaul network has a mixture of short (<20 km), medium (20 to 80 km), and long spans (80 to 120 km) that are optimally supported by single-stage and dual-stage EDFAs alone and dual-stage EDFAs with Raman amplifiers, respectively. Post-amplifiers are used as needed on longer spans or to compensate for the increased loss of a multidegree reconfigurable optical add/drop multiplexer (ROADM) versus a fixed OADM.

A combination of multidegree ROADMs, two-degree ROADMs, and fixed OADMs addresses all optical termination requirements. Multidegree ROADMs are typically used at fiber junction points, because they allow optical switching between multiple network fibers, as well as local add/drop. Two-degree ROADMs also enable any wavelength to be reconfigured between pass-through and add/drop without any service interruption, but they do not provide the inter-ring switching function. Fixed OADMs adequately service low-capacity add/drop sites.

Traffic passing from one ring to another may do so with or without OEO regeneration. Optical switching enables cost reductions, particularly with a larger number of pass-through wavelengths and at high data rates. OEO regeneration makes economical sense if the number of wavelengths switched between the core and access ring is very low, and even more so, if the shorter-reach (and therefore less expensive) lasers can be used on the access rings. For example, as illustrated in Figure 1, a single-span access link can use CWDM wavelengths with OEO regeneration at the interconnection point with the core DWDM ring.

A modern optical network is required to aggregate, transport, and switch tributary traffic with different protocols, data rates, and protection requirements. Without over-engineering, this can be accomplished using a minimal set of transponders and muxponders operating at 2.5-, 10-, and 40-Gbit/sec data rates.Th 211921

Figure 2. The ability to mix ROADM types on the same ring not only increases efficiency but lowers cost.

At 10-Gbit/sec modulation rates, a multihaul network would combine electroabsorptive modulated EML transponders for shorter reach (or where a lower number of nodes is traversed) and, where necessary, longer-reach transponders based on Mach-Zehnder modulated lasers. These latter modules would contain advanced features such as electronic dispersion compensation to mitigate accumulated chromatic dispersion, polarization-mode dispersion, and distortions caused by nonlinear fiber effects. In a mixed 2.5/10-Gbit/sec network, the 2.5-Gbit/sec wavelengths can take advantage of the DCMs deployed for the 10-Gbit/sec wavelengths; therefore, a 2.5-Gbit/sec transponder with a DML would be sufficient.

Managing all of the combined technologies and network segments can be facilitated by control plane software, such as GMPLS, to manage networks without regard to the traditional segmented boundaries of optical networks. GMPLS has simplified the provisioning, routing, management, and alarming of services, while taking advantage of new technologies such as ROADMs and optical switching.

The requirements of a multihaul network would appear to demand the purchase of a variety of different platforms or, worse yet, force-fitting existing single-purpose systems into applications for which they are ill suited. However, modular DWDM systems have reached the market that enable carriers to flexibly populate each network element with the capabilities and technologies that most closely match their requirements at each node or hub. This optimized DWDM system supports multiple options for each network function-amplifier, OADM, and transponders-on the same fiber span, with the ability to mix and match circuit packs as needed to build a reliable yet cost-effective DWDM transport network.

The network-level savings enabled by such multihaul DWDM systems are illustrated by a simple example in which the cost of the same ring is calculated in four cases, each with a different OADM option, while keeping all other equipment the same. A 12-node ring is assumed, with evenly distributed traffic termination among the OADM nodes of 20 or 40 wavelengths per fiber at 10 Gbits/sec. The ring is part of a larger network, and two of the nodes on this ring are hubs where 50% of the traffic is transferred to another DWDM ring.

The network cost for this ring example is calculated in the following cases:

• A, with fixed OADMs at every node.
• B, with two-degree ROADMs at every node.
• C, with four-degree ROADMs at every node, but nodes only equipped for two degrees, leaving the option of adding optical inter-ring switching in-service at a later date.
• D, with a mixed network having six 2-degree ROADMs, four fixed OADMs, and two 4-degree ROADMs.

The results are shown in Figure 2. The case with fixed OADMs at all nodes has its low OADM cost offset by the added cost of OEO regeneration required after six spans, due to excessive power ripple penalty and optical signal-to-noise ratio degradation. All other cases do not require regeneration, due to the power equalization benefits of the ROADMs.

The cases without four-degree ROADMs at the hub nodes suffer from excessive OEO regeneration in traffic handoff between rings. Cases C and D assume a four-degree ROADM at the two hub sites to enable optical switching between rings. The results clearly indicate that capital savings are attained using different OADM types on the same ring to optimally design the network to an operator’s specific requirements.

The evolution of DWDM networks has led to a merger between previously distinct application domains into a single optical network and to a large variety of requirements among different networks. This has generated the need for a modular DWDM system with a common hardware and software platform, offering multiple options for each system component. Unified control plane software, such as GMPLS, has accelerated the unification of optical networks by simplifying end-to-end management. The resulting multihaul DWDM system can be tailored to match the immediate and future needs of a network operator, spanning the entire range of application domains and customer-specific requirements, in a reliable and cost-effective manner.

Roy Bowcutt is vice president, product management, and Sorin Tibuleac is director, product management, at Movaz Networks (

More in Network Design