Wavelength management for all-optical networks and the components that support this architecture will allow metro carriers to efficiently offer multiprotocol, bit-rate-independent services.
All-optical networks are coming. An unrelenting rise in data traffic is pushing carriers to create efficient, dynamic networks. Ad vances in optical solutions--inherently flexible and more cost-effective--will enable this rapid expansion.
Dense wavelength-division multiplexing (DWDM) technology is a recent example. After point-to-point DWDM systems became commonplace in long-haul networks, an evolution to meshed and ring-type DWDM systems in metropolitan area networks (MANs) is underway. An architecture that will allow carriers to control and manage these wavelengths is mandatory in the competitive metro environment.
The movement toward all optics holds many challenges. A key step in this evolution is to define and create the optical components that will efficiently manage and operate these networks. These components include add/drop multiplexers (ADMs) and demultiplexers, attenuators, channel equalizers, and optical switches.
Until now, product advances within the optical data network primarily were targeted at the public-switched segment of the network or the long haul. Examples include 2.5-Gbit/sec and 10-Gbit/sec transmission rates and DWDM.
Implementation of point-to-point DWDM systems in long-haul networks was initially driven by fiber exhaust. To avoid the cost of deploying new fiber spans in excess of 400 km, the individual wavelengths served as virtual fiber. Virtual fiber removed the bottleneck in the long haul and provided other options, such as upgrading to denser channel spacing or opting for a larger number of channels by extending to the L-band (1565 nm to 1600nm). Today, the long haul is well defined. Technology approaches to the MAN, however, remain the focus of intense debate.
On the market for over two years, point-to-point DWDM systems have had limited commercial success within metro networks. While the threat of eventual fiber exhaust exists in the metro, the market drivers for DWDM systems in this segment remain distinct from those of long-haul networks.
The challenge for carriers is to create a metro network that is profitable and efficient. These networks must handle multiple services--business-to-business, business to the wide area network (WAN), video, and typical Internet protocol (IP) and voice traffic. And metro networks must handle these services transparently, which requires support for data rates that range from 155 Mbits/sec to 10 Gbits/sec and protocols that include such diverse formats as SONET/SDH, IP, Asynchronous Transfer Mode (ATM), enterprise service connection (Escon), Fibre Channel, and Gigabit Ethernet. By handling multiple protocols at varying speeds, the network is more efficient and can produce additional revenue.
High reliability and scalability are critical within the metro segment, given the amount of traffic handled by the network. Elaborate networks cannot be de fined, implemented, and expected to pay for themselves overnight, however. Instead, the technology must be flexible enough to support a gradual evolution of the network.
Thus, a new direction is underway to provide bandwidth-management capabilities in the metro and facilitate a conversion between the WAN and the local area network. The particular needs of the metro and the promise of an all-optical network are fueling the concept of wavelength management. This architecture allows provisioning, restoration, and bandwidth management to take place in the optical layer (see Figure).
The quest for protocol- and data-rate-independent architectures in the metro mandates wavelength-management capabilities. The components required to implement wavelength management include programmable, optical ADMs; optical switching for restoration of the optical layer; protocol- and data-rate-independent wavelength provisioning; and tunable sources such as tunable-laser transmitters.
Within a ring architecture, wavelength ADMs are fixed. Remote configuration of wavelengths is just beginning with the introduction of wavelength-tunable sources. Two technologies will accelerate the development of dynamic and configurable optical rings-micro-electromechanical systems (MEMS) and silicon optical bench, which incorporates waveguide with optical-switch structures. The development of these technologies will pave the way for greater remote control to govern the direction of wavelengths on the ring, adjust the optical power of those wavelengths--or at minimum, to monitor performance.
Wavelength management will result in a more-efficient network for metro carriers, all the way down to the access level. Development of the components that will support this architecture--and defining the features of those devices--will allow all-optical technology to take hold. Key components include add/drop multiplexers and demultiplexers, attenuators, channel equalizers, and optical switches.
In the all-optical network, today's SONET ADMs, which convert optical signals to electrical signals and back again, will evolve into optical ADMs. Each time a wavelength is added or dropped within an optical ring, the power fluctuates. Thus, optical amplifiers that can monitor these fluctuations and make adjustments are required.
This capability is also important at the receiver end, where a circuit can be easily overloaded. An optical budget is typically on the order of 30 to 32 dB. When a wavelength is dropped, the optical budget can fall by as much as 20 dB, depending on where the "drop" is located on the ring. Within a 32-dB budget, a slip to 20 dB by dropping one wavelength can cause oversaturation of the receiver. To compensate, an optical attenuator either in the amplifier, receiver, or source is required. Similarly, in an optically amplified system, power fluctuations due to dynamic optical add/drop multiplexing requires a more complex optical amplifier.
Gain flatness is accomplished on a dynamic basis to avoid large spikes of power from adding wavelengths and big reductions of power from dropping wavelengths. Within a 64-channel system, for example, dropping 10 to 12 channels requires that the optical power be maintained at a constant level, so gain saturation won't occur in the optical amplifier.
Because the exact location of where wavelengths will be added or dropped is unknown, it is critical to maintain power levels within the optical power budget of the ring. Thus, channel equalization performed in a centralized location is sometimes advantageous. One place to control power is inside the optical amplifier. The gain channel power level, or the attenuation of individual channels, can be controlled either inside or outside the optical amplifier. To drop 10 wavelengths in a 64-channel system, for example, each wavelength may have a variable optical attentuator (VOA). A VOA on each channel with a channel equalizer placed inside an optical amplifier can control fluctuations; another option is a component that will monitor and adjust the power levels of all the add/drop channels at once. As network architectures mature, specifications will be developed.
Where the gain or channel equalization takes place will be determined as all-optical networks evolve. To maintain a constant level, wavelength channel monitoring--a consistent monitoring of the signal-to-noise ratio (SNR) of each channel--is necessary. The SNR is controlled at the receiver, or source, either by a VOA in each channel or by a single channel-equalizing device.
Currently, the adding or dropping of signals is accomplished with bulk optics. Individual components such as optical multiplexers, optical demultiplexers, 2x2 optical switches, and VOAs are used to "build" a programmable optical add/drop network element. Unfortunately, this solution is costly and labor intensive.
By driving integration down to the component level, it is possible to get away from bulk optics with a silicon optical bench. A silicon array waveguide is used with optical multiplexers and demultiplexers. In this case, optical switching is done with thermal-optic switches.
Interconnectivity between rings for provisioning and restoration is also necessary. Today, this function is performed at the electrical level with digital crossconnects at lower bit rates. In the future, it can take place at higher bit rates in the optical layer. In theory, it can be done with optical switching. With such systems, the need to do protocol monitoring and conversions is minimized--as is the necessity for racks of electrical equipment. The challenge then becomes the provision of reconfigurable optical switching-being able to take any input and reconfiguring that input port to any output port.
The bulk optics used today in programmed optical add/drop network elements can serve as prototype systems for reconfigurable optical switches. Since these "elements" are large and costly to construct, miniaturization offers an advantage. Silicon optical-bench technology combined with thermal-optic switches points to the possibility of smaller-scale switches. In this case, miniaturization will provide cost-effective elements that can meet performance objectives and system requirements as carriers evolve their infrastructures to all-optical networks.
The optical layer provides a platform for transparency between protocols and supports network operation at varied data rates. It also will allow service providers to re-route traffic efficiently throughout the network. Today, traffic re-routing is performed by electrical and digital crossconnects at low data rates within an electrical switching fabric. With an optical-switching fabric, re-routing can occur at higher data rates.
Today, the fiber-optics industry is defining the products that will support all-optical networks. Several products will eventually move under the wavelength-management umbrella as these networks rapidly evolve. Exactly when the evolution to all-optical networks will take place is the subject of many debates. It is expected, however, that several network elements on the market this year will require corresponding wavelength-management products to operate.
To understand the potential of wavelength management, just look at the rapid evolution of WDM. Initial expectations were that WDM would support 4- or 8-channel systems. Today, it's approaching 128 channels. The skeptics at one time also agreed that 50-GHz spacing between wavelengths could never take place. Those same naysayers now build 50-GHz WDM point-to-point systems.
Data traffic is doubling each year. While the timing is open to debate, the industry's move from fixed to dynamic components is a necessary transition for the evolution to multiprotocol, bit-rate-independent metropolitan networks.
Joseph J. Calvitti is a marketing development manager at Lucent Technologies Microelectronics Group (Breinigsville, PA).