Cost-optimized designs for metro regional optical networks
By PER B. HANSEN, Photuris--The metro regional market remains underserved today because of a poor understanding of its system requirements. Unlike access and long-haul networks, metro regional networks--loosely defined as networks with circumferences between 50 and 500 km--are characterized by high variability in network topologies, span lengths, and losses.
By PER B. HANSEN, Photuris--The metro regional market remains underserved today because of a poor understanding of its system requirements. Unlike access and long-haul networks, metro regional networks--loosely defined as networks with circumferences between 50 and 500 km--are characterized by high variability in network topologies, span lengths, and losses. Span lengths can vary from a few hundred meters to more than 100 km. Losses can range from less than 1 dB to more than 25 dB, or anywhere in between. A variety of fiber, and even several types within one system, is also par for the course. This "fuzzy" design target is a challenge for equipment manufacturers, but also an opportunity for deriving value from optical technology trends and novel network designs. Recent advances in optical technologies and cost pressures, especially on components, are finally making systems optimized for the metro regional market viable.
Obvious prerequisites are lower cost, lower loss, multifunctional devices and subsystems, which enable the greatest number of system-level features at the least cost. Less obvious methods for lowering system cost include the choice and use of various optical amplifier designs and dispersion compensation. When properly applied to metro regional topologies, these design decisions offer significant cost savings.
A network's overall utility is defined as the serviceable capacity demand relative to the required capital investment. Overall utility, and how well capital investment is aligned with revenue generation, are the fundamental metrics a network operator considers when making equipment purchase decisions. Maximizing utility leads to a mandate for single-wavelength granularity in a broad sense. Essentially, all service and wavelength options must be available, independent of provisioned services. Restrictions on line data rates, the range of protection schemes, and network connectivity including wavelength reuse all affect a network's overall utility.
Previous metro regional network installations were limited by the available technology. Installed non-regenerated networks, for example, exhibit single-digit node counts because the utility of banded network solutions drops quickly as nodes increase.
Potential cost savings from minimizing regeneration, however, is driving networks towards higher node counts of 8-12, and even 16 nodes, allowing for consolidation of smaller rings (see Figure 1). The move to one large ring also lowers cost on the operations side.
A modular architecture, which aligns cost and revenue, helps cash flow management. Ideally, the operator would like to pay only for lit wavelengths, and even then, not until the wavelengths are actually provisioned. Modularity applies to several dimensions, including capacity as represented by the number of wavelengths, and topology metrics such as span length and loss.
Optimal system design
Large variations in span length and loss in metro regional networks have several implications for optimized system design. The typical distances and associated fiber losses make optical amplification an absolute must for metro regional networks. Variable-gain, low-noise amplifiers, which adjust gain to match loss without costly, error-prone manual procedures, are also mandatory.
While variable-gain amplifiers address the problem of adding loss to a span to match a fixed-gain amplifier and therefore offer savings in terms of less labor-intensive installation, they do not reduce the cost of the amplifier itself. The hardware cost may be even higher. Savings from less expensive hardware is only realized when a range of amplifiers targeting different span losses is used. Smaller, more cost-effective amplifiers can be applied in spans with relatively low loss.
As an integral feature of node design, power control is either provided on a broadband or per-wavelength basis. Even small gain ripples from gain-flattened amplifiers -- the most common approach for broadband power control-- accumulate to significant power deviations in networks supporting up to 16 nodes. Channels disadvantaged by insufficient gain will, of course, suffer performance degradations.
Gain spectrums may change significantly, depending on the number of amplified channels. As a result, single channel performance may vary as other channels are provisioned, or taken out of service.
Networks supporting large node counts benefit from dynamic per-channel power control, since this decouples the performance of a wavelength from other provisioned wavelengths and eliminates systemic residual gain ripples. Over many nodes, these ripples may lead to significant power droop for some channels. The benefits of an active power control scheme thus apply both to system performance and robustness.
Dispersion compensation and optical amplifier allocation
Contrary to the dispersion maps of long-haul systems, the less stringent requirements in metro regional networks are achieved by appropriately limiting signal launch powers. An allowable window for the accumulated dispersion requires fewer, less customized dispersion-compensation modules (DCMs).
Less stringent requirements also offer the opportunity to locate DCMs so that associated loss is compensated in the most cost-efficient way. This is accomplished by allowing DCMs to share in span power budgets by locating the modules so that meeting the signal-to-noise requirements for all possible network connections requires the lowest additional cost from amplification (see Figure 2). For simplicity, only the clockwise traffic and associated equipment are discussed here. In this case, two short low-loss spans provide ample power budget for two DCMs. The impact of their insertion loss on service performance is insignificant, eliminating the need for a more costly amplifier to accommodate dispersion compensation between stages.
Over-compensation may lead to an even lower total system cost for some topologies. Consider a similar case with span lengths such that the implementation shown in Figure 2a satisfies the dispersion rules with only one DCM in the amplifier of node C (not two, as shown in the Figure). Eliminating the pre-compensation in node C, which compensates span C-A still results in a need for two DCMs, one in each of the spans, A-B and B-C (see Figure 2b). This may come about because clockwise connections from node B to node A and node C to node B both benefit from one DCM if it is associated with the C-A span, whereas two DCMs are required if no DCMs are associated with that span. Although the number of DCMs increases by one, it may still be the lower-cost solution, depending on the cost difference for the two amplifier codes.
Optimal allocation of DCMs and amplifiers is not trivial because of span loss limitations determined by each amplifier code and DCM loss, as well as the impact of both components on the optical signal-to-noise ratio. Spreadsheet representations of engineering rules are inadequate, and invariably lead to excessively conservative equipment and cost estimates. Software applications can allocate equipment according to the network topology and span characteristics, and take advantage of numerical methods to provide an optimized solution despite the huge number of variables to be considered.
Network cost efficiency
The cost efficiency of a metro regional system depends on the richness of amplifier and DCM options. Although one variable-gain amplifier code with a sufficiently large dynamic range to accommodate the full-range of span losses may sound appealing from the perspective of equipment sparing and product line management, it is synonymous with employing a very capable and, therefore, costlier amplifier, even for very short spans with insignificant loss. A range of amplifier codes allows the network planner to deploy a less expensive solution when spans are not very challenging.
How many amplifier codes are required? Although the number is probably in the single digits, there is obviously no easy answer. In general, the range of capabilities relative to the range of cost determines the optimum set of codes. Not surprisingly, neither extreme is attractive. A very large number of amplifier codes is hard to manage and expensive to spare. A single code provides no options for cost savings and leads to less cost-efficient network installations.
Similar arguments can be made for the number of DCM codes. However, with dispersion-compensating fiber as the benchmark, the inherent cost is to a much higher degree proportional to performance (i.e. amount of dispersion compensation), simplifying the issue.
Modularity is by no means a novel concept, but it is critical for cost-efficient metro regional network architectures. As access to capital has diminished, operators increasingly look for solutions to minimize cost of ownership and let them delay investments until demands and associated revenues materialize. Given the variability in network topologies as well as network loading over time, cost-efficient architectures rely heavily on modularity. An appropriate range of amplifiers and DCMs address the issue of variability in reach, whereas modularity in the wavelength domain allows cost-efficient solutions to grow, starting with a single wavelength.
Per B. Hansen, director of business development with Photuris (Piscataway, NJ), is lead designer of the optical transport layer defining the system engineering rules. He was previously with Bell Labs, Lucent Technologies, where his areas of research included pulsed monolithic semiconductor lightsources, Raman and remotely pumped erbium-doped fiber amplifier system applications, ultra-long span repeaterless submarine systems, and high capacity long-haul WDM transmission systems.