During the past three years, some of the most significant advances in optical networking have been made in the areas of system reach and photonic or all-optical switching. The combination of increased reach and photonic mesh connectivity results in reach that can be applied to the (A-Z) traffic across a network, rather than just a single link. The resulting 'transparent network' extends the concept of regenerator elimination from the line to the nodes.
The elimination of unnecessary intermediate optical-electrical-optical (OEO) interfaces between DWDM systems results in greater than 50% network cost reduction. However, this photonic mesh does not completely eliminate OEO interfaces, since signal degradation due to both distance and intermediate switching elements still results in some connections requiring reach extension or wavelength conversion at the photonic node. A close look at the A-Z traffic demand shows that unique start and endpoints drive a wide variety of reach requirements for individual wavelengths in a given network, a term we call 'network-derived' reach.
When technologies used in DWDM systems, such as EDFAs, Raman amplifiers, and dispersion compensators are combined with different transmission fiber properties, channel performance is affected in different ways across the DWDM spectrum. As a result of these spectrally dependent effects, individual wavelengths have different reach capabilities from one another.
This causes problems in traditional DWDM setups that focus on reducing line regenerators between endpoints of the system. The idea there is to ensure that all wavelengths reach the maximum limit of the system, but this design typically lowers the average reach of the system to accommodate the poorest performing wavelength. This compromises the performance potential of the best channels.
Because transparency allows wavelengths to extend beyond the limits of a single link, it presents an opportunity to remove that compromise, by dynamically assigning wavelengths based on connection-length requirements. This way, the overall system capacity and reach can be used more efficiently. Longer connections would use better performing wavelengths, and therefore achieve longer effective reach. Naturally, shorter connections use the poorer performing wavelengths. The result is a net reduction in the number of regenerators across the network.
The "agile reach" model introduces additional requirements on the system. It requires that the system be equipped with full wavelength tunability to enable unconstrained selection of a wavelength upon setup of the connection. Furthermore, intelligent control and routing systems are required to maintain information about network attributes such as performance attributes and interaction (due to system fill) allowing for the intelligent assignment of future wavelength set-up requests. Due to the variability of a network's characteristics, it is imperative that the system be capable of dynamically measuring and updating this information to ensure optimal wavelength selection in all cases.
By allocating the A-Z connection distance to a wavelength of comparable reach performance, the overall system capacity and reach can be used more efficiently. This method allows longer demands to use better performing wavelengths, and therefore achieve longer effective reach. Naturally, shorter A-Z requirements utilize the poorer performing wavelengths. When the full performance range of the system can be fully utilized, the result is a net reduction in the number of regenerators across the network.
We examined the economic effects of mapping the natural reach distribution (agile reach) of a DWDM system to A-Z network demands.
A-Z network demand requirements vary depending on the network in question. The analysis conducted was based on a 20-node North American backbone network model. In order to reflect A-Z traffic requirements, a 'gravity model' was used. A base model was used reflecting 3 Terabit of aggregate traffic between a total of 110 nodes. The traffic from all 110 nodes was groomed onto the 20-node backbone utilizing sub-wavelength bandwidth management. This provided optimized 10 Gbit/second wavelength connections for each of the equipment models.
The focus of the analysis was to understand the effective reach increases made possible by matching demands to a wavelength distribution through agile reach. Three network architectures were modeled:
-Point-to-point DWDM equipment model with static reach. This is a traditional point-to-point system with a single reach for all wavelengths. This system requires OEO conversion at all intermediate junction sites, since the junctions define the endpoints of the point-to-point system. This model assumes "line regeneration" (between junctions) is eliminated by the reach capabilities.
-Transparent (Photonic Mesh) system with static reach. This is a transparent system with a single reach for all wavelengths. This system allows wavelengths to optically pass-through junction sites if the reach permits. Otherwise, OEO conversion is used at the appropriate intermediate junction.
-Transparent (Photonic Mesh) system, with agile reach. This model employs a varying reach distribution and intelligent control systems able to match connections to wavelengths with appropriate reach. OEO conversion is used at intermediate junction sites where needed to extend beyond the available wavelength reach.
A number of reach distributions were compared. In order to identify an appropriate target average reach, a network cost comparison of line technologies was completed. DWDM system reach comes as a trade- off of many factors including spectral efficiency (bits per second per fiber), line amplifier and opto-electronics costs. Network-defined reach dictates the appropriate balance of system reach and capacity, as it varies depending on traffic pattern and physical topology.
For the study, four reach and capacity combinations were compared on a transparent system using the traffic and network model. The optimal reach for the network was around 3000 km (average reach), with 1 Terabit/s line capacity.
Based on the analysis, the focus of the model assumed an average reach of around 3000 km. For the network architecture, we considered four hypothetical reach distributions. A system with 3000 km average reach, if applied to a static network or a point-to-point network would result in perhaps a flat 2400 km reach if pre-emphasis was applied to recover performance of the poorest performing wavelengths. Thus, a flat 2400 km reach was applied to the point-to-point and the transparent-static architectures.
The same component transmission technology could provide an approximate agile reach profile of about 3000 km average reach with linear variation over wavelength of 50% of that average reach. That is, the shortest reach and longest reach wavelengths have 1500 km and 4500 km reach, respectively. This reach distribution was applied to a transparent network with dynamic reach matching capability. Finally, distributions with average reach of 2400 km and variation of 25% and 50% were applied to the dynamic transparent network to reflect sensitivities of a cost-reduced line system with lower average reach.
In the results of the routing and reach analysis showed that the average optical reach was achieved when connections were mapped across each network equipment model with the selected wavelength reach distributions.
Also, the relative (nodal) regenerator count provided a stronger indicator of relative cost of the networks. Traffic and equipment models considered, transparent networks offered 52% regenerator savings for the network over point-to-point systems, by enabling the network to eliminate OEO regeneration through transparent switching at junction sites. Introducing dynamic reach provided an additional 51% reduction in regenerators over the static reach model by increasing the average reach and matching individual connection reach requirements to the inherent wavelength reach distribution in the system. Altogether 76% of all intermediate regenerators can be removed by introducing a transparent network with agile reach.
When the average reach was held at 2400 km and the inherent reach variation in the system grew from 0% to 50%, the average achieved reach grew significantly. This implies that dynamic reach matching in a transparent network may allow reduction in both the line system costs and regenerator costs simultaneously. It also suggests that conventional optical system performance metrics typically limited by the reach of the poorest wavelength are not appropriate for transparent networks where greater reach variation and thus, lower minimum reach may imply greater average achieved performance.
Matching wavelengths to connection length can drastically reduce network regeneration requirements and cost. For a static reach of 2400 km, agile reach for a similar system provides greater than 50% regenerator savings. Network cost benefits from reduced regeneration, and can be further optimized by removing some line system components, while maintaining the same regeneration reduction using dynamic reach.
To take advantage of dynamic reach, a system must allow dynamic provisioning of wavelengths mapped to demands at turn-up time, facilitated by a control system that is "network-aware" of the transmission characteristics of the composite links.
See related chart