The path for wavelength switching

Many technologies have been introduced as enablers of optical-wavelength switching, each with its own set of advantages and disadvantages. Wavelengh-switching technologies are embedded within standalone network elements, such as photonic cross-connects, or in subsystems such as a remotely configurable optical add/drop multiplexer (OADM). Although these technologies have been successfully demonstrated, some noteworthy legacy issues within current optical networks have hindered or even temporarily shelved their commercialization success as of today.

Today's optical networks switch wavelengths using fixed OADMs that are deployed selectively within the optical line between the terminals. The OADMs are chosen in a predetermined manner so as to add or drop certain wavelengths or bands while allowing others to pass through as "express" wavelengths.

Although this setup does yield a robust and cost-effective solution, it does not allow for dynamic switching, since the switching device itself, a passive OADM, is fixed.

Added and dropped wavelengths are predetermined based on forecasted traffic demands that inevitably change, resulting in possible suboptimal network designs. However, for the most part, the passive OADM serves carriers very well, especially with the more moderate and manageable projected traffic growth rates of today as compared with the boom years.

Two obvious methods exist to enable the switching of wavelengths. Passive OADMs can be used with tunable terminal transmitters, but this calls for redesigned mux/demux filters. Alternatively, fixed transmitters coupled with the passive mux/demux filters already deployed could be used, with wavelength switching occurring within a dynamic provisionable device along the optical line.

The latter is a more sensible choice, since it involves adding the cost and complexity only to the switching element itself, rather than to all of the terminal transmitters and mux/demux filters. It also allows for a smoother migration toward a wavelength-switched network, since the existing fixed transmitters and mux/demux filters could still be used. This would leverage rather than strand the huge base of installed equipment.

Indeed, the issues to be addressed in moving to wavelength switching are not so much related to optical-switching technologies themselves, but to the manner in which currently deployed networks were originally designed and the resulting migration issues.

The majority of optical networks now in use are passive in nature from a wavelength-switching perspective. In other words, the optical path of a given wavelength is predetermined based on the current and projected network traffic characteristics. The modules that actually dictate a wavelength's chosen path (mux/demux filters, transmitter wavelengths and OADMs) are all predetermined (fixed). Tunable lasers are now being deployed, but primarily for reasons such as equipment sparing or increased flexibility for multiport optical-interface cards.

Integration hurdles

The effective implementation of wavelength switching along diverse optical paths will require overcoming certain impediments related to the enormous installed network base. In today's financially constrained telecom industry, few carriers will choose green-field network builds based on yet-unproven technologies if their existing infrastructure is not fully utilized. A less risky strategy that leverages existing network investments is preferred. But there are some significant issues surrounding the integration of wavelength switching into existing networks.

• Equalization: Network robustness dictates that optical lines be equalized, with each propagating wavelength exhibiting a relatively equal optical signal-to-noise ratio at their receivers. Due to the nonlinear optical characteristics of erbium-doped fiber amplifiers, this involves adjustments at the transmitter output powers or in the amplifiers in the optical line. The quick and dynamic adding/dropping of switched wavelengths in and out of a given optical line requires much more advanced and complex optical-control mechanisms than exist today.

• Hardware optimization: Switching wavelengths between any two network locations assumes that the required receivers have already been deployed at all of the allowable termination points. This results in many unused-yet paid for-receivers, incurring wasted capital expenditures that are not generating much-needed revenue. Flexibility vs. cost in optically switched networks will be a significant challenge, especially in today's economically constrained telecom industry.

• Network management systems: The NMS would be expected to fully monitor and control all possible switching paths from end to end in real-time, in both a proactive and a reactive manner. To do so would require a major overhaul of how the NMS is designed and deployed. For starters, the updated NMS must avoid switching a given wavelength onto an optical line that already has that specific wavelength in service.

• Traffic patterns: To cost-effectively optimize a given optical network, the highest line rate available is used to minimize the number of transmitters and receivers. Currently, 10 Gbits/second is the highest line rate deployed, making it the optical granularity between any paired network ingress/egress points. Most end users do not require 10 Gbits/s of traffic, resulting in numerous customers simultaneously sharing the same given 10-Gbit/s wavelength. To ensure optimization of wavelength use, subtending electrical bandwidth-grooming switches are currently used.

• Interoperability: As networks become increasingly "optical," they also become more proprietary due to the lack of standards at the optical layer. Although wavelength grids are standardized, how they are monitored and managed is a problem in terms of optical midspan. More global standards are required to further advance the commercialization of wavelength switching while at the same time lowering the production costs of components and systems.

• Protection schemes: Many customers currently share a given 10-Gbit/s wavelength and thus also share the same class of service from an optical-switching protection viewpoint. In many cases, this makes sense, but it does not allow carriers to offer differentiated classes of services. Bandwidth management and protection at the electrical layer allow for customers to share the same 10-Gbit/s wavelength, albeit with different protection schemes at different costs.

Does this mean that optical switching is not possible? Certainly not. The question is not if, but when and how, it will be deployed.

Although there are many applications conducive to wavelength switching, the main ones are related to dynamic on-demand wavelength services, line-rate granularity bandwidth management and enhanced protection schemes.

On-demand dynamic-wavelength switching allows wavelengths from one network location to be dynamically switched to another network location based on specific demands. Examples range from video fed from a specific sports event to remote data archiving of storage-area networks. Bandwidth management at the wavelength granularity allows wavelengths to be routed purely in the optical domain for customers sharing the same network ingress and egress points. Enhanced protection schemes would dynamically route entire wavelengths around problem points within a given network infrastructure.

Alternatively, these three primary applications are being cost-effectively achieved at the electrical layer using optical-electrical-optical (OEO) cross-connects and next-generation service switches. For wavelength switching to take hold in the marketplace, it would have to improve, expand or lower the total cost of traffic transport when compared with existing electrical solutions. Regardless of the specific target application, the cost-effective and relatively painless integration into today's optical networks is absolutely mandatory and is expected by carriers.

Popularity grows

Significant strides in the area of wavelength switching have occurred over the past few years as a result of the influx of venture capital into the optical-networking industry. Many technologies have appeared and adapted, while others have simply withered away. Wavelength switching becomes increasingly desirable as the granularity of end-user traffic approaches the line rate of the network-currently 10 Gbits/s.

Most customer traffic is less than 10 Gbits/s in granularity, however, making wavelength switching for bandwidth management less enticing. This reality mandates the use of OEO switches that perform bandwidth management and protection switching within the same network element.

For wavelength switching to achieve significant commercial success, it must first find a cost-effective migration path within the designs of existing optical networks while at the same time lowering the overall networking cost.

See related chart