Over the past two years, advances in ASIC design, serial interfaces, optics and ultralong-haul transmission components have produced key enabling technologies that can be leveraged into higher-capacity, more cost-effective OEO switches. Using these technology advances, equipment suppliers can now build single-bay systems that switch multiterabits per second of capacity, which is five times greater than some current systems; support more than 1,000 optical ports; and provide more flexible system configurations. This increased capacity, port density and flexibility radically improves per-port economics.

Among those key OEO technologies, ASIC processes have advanced considerably, with 130-nm and, more recently, 90-nm fabrication technology. This allows significantly more functionality to be packed into devices that have the same size and power as in current switching systems. In particular, framing devices can now integrate multiple I/O channels with increased overhead processing capabilities, and switching devices offer greater capacity at finer granularities.

Improving capacity

As the capacity of the framing and switching components increases, it follows that the interconnection of those devices will require improved transmission capacity. Further, as more functionality is added to components, device pin counts and pin spacing become a growing concern. Here, newer serialization technologies are delivering higher transmission clock speeds, in the range of 5 GHz to 10 GHz, across current backplane technologies. In addition, built-in equalization serves to cut device pin count demands.

In addition, pluggable optics modules can accommodate a range of optical interfaces on the same host electronics. In today's systems, I/O cards on a given OEO system tend to be customized to a narrow range of options for line rate, frequency or reach. But the advent of industry-standard form factors for pluggable optical modules, in particular SFP and XFP (small-form pluggable) modules, allows for host I/O cards that have improved port densities with an increased range of rate, frequency and reach.

Because these optical modules are standardized and gain more acceptance in the vendor community, it follows that a carrier can maintain a common pool of spares that applies to a broader range of installed equipment. But even the SFP and XFP optical interfaces have yet to deliver the optical characteristics required for dense wavelength-division multiplexing (DWDM) ultralong-haul transmissions of 1,500 kilometers or more. As a result, for these applications, it is still necessary to build optical modules from discrete components.

Other significant technology advances have led to reductions in both size and power for such components as uncooled distributed feedback lasers, modulators and frequency-locked lasers. Thus, it is now possible to integrate optics for the most sophisticated DWDM transmission applications directly into the OEO switching system.

Improvements in capacity and transmission distances of DWDM systems coupled with new OO switching architectures and switch component technologies have opened the door to practical, cost-effective OO switches.

All-optical switching is conceptually different from OEO switching because it is fundamentally analog, not digital, in nature. This means that the OO switch must be embedded in an optical line system, and its optical properties directly affect the transmission. The ideal switch will manipulate individual wavelengths into and out of a transmission path without affecting the unswitched portion of the spectrum.

A "broadcast and select" optical switch architecture provides a low-cost approach that enjoys both operational advantages and minimizes optical impairments. The broadcast part of such a switch architecture is accomplished with passive splitters, while the select part is enabled by reconfigurable wavelength blockers (WBs), a key technology that makes OO switching practical. A WB is a component that integrates wavelength-level multiplexer/demultiplexer (mux/demux) capability with the ability to selectively block wavelengths from the transmission path.

There are two distinct applications for the mux/demux technology. The first application is to enable wavelength-specific switching by first separating the wavelengths so they can be individually manipulated, and then recombining them. For this application, a high premium is put on the ability to integrate the mux/demux function with the switch fabric. It is also important that the mux/demux function be cascadable, as a signal could traverse many switches as it travels through a network.

The second application for mux/demux technology is to allow local add/drop traffic at a switch site. Ideally, the mux/demux technology for this application would allow arbitrary wavelength adds and drops with a minimal parts count. For example, an application that drops eight wavelengths spaced across the spectrum should require only those parts needed to drop the eight channels and not those for the whole spectrum.

Two key performance metrics of the WB are its ability to remove as much of the blocked signal as possible, and the ability to minimally disturb the express path. Incomplete signal removal causes degradation of channels added subsequently through a mechanism known as coherent interference, essentially a form of in-band crosstalk. The degree to which any given signal is blocked is measured by an extinction ratio, defined as the signal level at the output of the blocker normalized by the level at the input.

Plugging the gaps

Available technologies for wave blocking are microelectromechanical systems mirror and liquid-crystal (LC) arrays. MEMS mirrors typically give the best extinction ratio, but leave gaps in the transmission spectrum. A better choice is LC arrays, which can be fabricated with a very small "dead zone" that minimizes the gaps in the transmission spectrum. These gaps are undesirable because they widen as the switches are cascaded, resulting in the narrowing of the channel passbands. At the system level, narrow passbands reduce the capability of the line system to carry a variety of data rates, channel spacings and modulation formats.

As higher-capacity optical switching systems are implemented to lower the cost per bit, as well as space and power requirements, it is important to consider the economic impact of the associated software. Software plays the critical roles of accurately maintaining inventory of optical resources, properly allocating them to services, monitoring them for faults and providing protection for service assurance.

Without the ability to maintain accurate network inventories or optimally allocate network resources during service fulfillment, poor network utilization will result, driving up the costs to meet service demands. Also, without the ability to accurately monitor and protect resources, labor costs for troubleshooting rise and revenue is lost to service-level agreement penalties. In short, without the benefits of highly functional, comprehensive software, the economic gains promised by the higher-capacity optical systems will be diminished by poor utilization and poor productivity.

Jeff Livas is chief technologist for the core transport group at Ciena Corp. (Linthicum, Md.); Tad Hofmeister is a senior principal engineer and Karl Horne is technology director, both in Ciena's core networking products group.

See related chart

See related chart

See related chart