Keeping the promise of optical switching for metro networks

June 1, 2001
Metro Networks

Optical switches hold the key to building reliable, high-performance, all-optical metro networks that upgrade easily and inexpensively; but ultimate success depends on making the right component choices.

ABE QUELLER, Lynx Photonic Networks

Metropolitan networks are defined as the segment of the transport network that is deployed in cities. They interconnect users within the same geographic area and serve as a bridge between access networks and the regional/long-haul network. These networks were originally constructed of fiber rings and run for distances of 150 to 200 km. The rings are interconnected via crossconnects, while add/drop multiplexers provide network access to customers.

Metro networks were primarily intended to carry voice traffic but were adapted over time to carry an increasing amount of data, encoded in time-division multiplexing (TDM) format. Increasing demands for higher bandwidth, and a change in the makeup of the traffic, have forced carriers to adopt new transport technologies based on WDM and, more recently, DWDM. Metro networks have unique characteristics and present unique trends, which require distinctive solutions.

Metro networks have a lower capacity than their long-haul counterparts and, therefore, contain a smaller number of optical-fiber strands. Moreover, metro networks have fewer channels within each strand, with the average network containing 16-40 channels per fiber. Although metro networks run for shorter distances, they have a far greater number of nodes and switch points.

Metro networks were built to carry voice traffic, for which the SONET/SDH protocol is most suitable. In recent years, due to the introduction of the Internet, these networks carry large amounts of data, transported over various protocols. The SONET/SDH mechanism is considered less suitable for this increasing flood of data traffic.
Figure 1. Today's metropolitan-area networks are being increasingly pressed to move from voice-only to voice and data services, while data rates are climbing from OC-3 to OC-12 to OC-48. In prevailing ring topologies, optical-electrical-optical crossconnects between rings are limiting implementation of newer, faster services.

While long-haul networks are designed to efficiently carry bulk traffic from one point to another, metro networks have to meet a different set of objectives. These networks must support quick provisioning and frequent reconfiguration to meet customers' needs. They also need to support a large number of applications and various high-margin services offered by the carriers.

Several trends in the behavior and characteristics of metro networks have helped shape their architecture and influence the design of the underlying technology. Increasingly, change is a fact of life for a metro network. In particular, metropolitan transport components are being pressed to provide higher data rates and channel capacity as well as accommodate new services and data protocols. While OC-3 (155-Mbit/sec) rates were used just a few years ago, today's networks boast data rates of OC-12 (622 Mbits/sec), OC-48 (2.5 Gbits/sec), and higher. Traditional metro networks, which use electrical-optical-electrical (OEO) technology at their switch points, have reached their practical limitation (see Figure 1).

Along with the advance in data rates, new protocols are being introduced to support new applications. The popularity of the Internet has increased the use of TCP/IP (transmission control protocol/Internet Protocol). Multiprotocol Label Switching (MPLS), considered by many as an enhancement of TCP/IP, is increasing in popularity and will undoubtedly play a major role in future metro networks.

Localization of metro traffic, due partly to the use of storage-area networks (SANs) and data caching, promoted the use of Fibre Channel and Gigabit Ethernet protocols in the metro network. Cell-based protocols, such as ATM, are frequently used, as well.

With these changes in link data rates and protocols, carriers must upgrade their networks quite frequently. These frequent and expensive overhauls to the network have prompted network architects to look for futureproof technologies such as all-optical networks that are transparent to data rates, protocols, and content.

Although SONET/SDH rings feature robust fault recovery, they are not scalable. They utilize a single channel per fiber and operate only up to certain speeds. Additionally, the ring topology, combined with outdated management tools, doesn't support fast provisioning and reconfiguration-considered critical to the carriers. These deficiencies are prompting network designers to migrate from rings to a mesh topology, utilizing DWDM links.

Increased reliance on enterprise networks and various network services has placed greater demand for network availability. Indeed, for this reason, meeting the much-quoted five-nines (99.999%) reliability standard in metro networks has become increasingly critical.

These trends-higher bandwidth, channel capacity, new protocols and services-beg for an all-optical network that can accommodate them easily. The main strength of all-optical network nodes, with no translation between optical and electronic signals, is their transparency to changes in network content, bandwidth, channel capacity, and protocol (see Figure 2).
Figure 2. Future metropolitan-area networks must be responsive to demands for higher bandwidth, new services, and the protocols to deliver them, providing impetus for DWDM mesh topologies with all-optical crossconnects and add/drop switching.

As part of the transition to all-optical networks, designers must eschew OEO switches for their all-optical counterparts. Fortunately, the optical-switch industry has recently yielded new and innovative technologies for the construction of optical crossconnects (OXCs), optical add/drop multiplexers (OADMs), and protection switches, allowing network architects to choose the right product for their application.

Switch-matrix size is key to selecting an optical switch. Designers should match switch matrix size to their application's needs to obtain optimal results. Using switches of incorrect size increases the total system cost and results in non-optimal performance. Most of the current metro networks require 6x6 and 8x8 switches. Future networks will require switch sizes of 16x16 and 32x32.

Another key factor to consider is optical power loss. Optical devices are characterized by their inherent loss, which may be attributed to changes in material through which light travels, internal geometries, material imperfections, and so on. As losses mount, signal quality deteriorates and optical amplifiers, as well as regenerators, are needed to maintain signal quality.

These devices, however, increase the network's cost, complexity, and likelihood of failure. Designers should opt for devices that exhibit minimum total loss and insensitivity to external variants. Signal quality is also affected by crosstalk on neighboring connections. Crosstalk values of -40 dB or less are considered ideal.

Reliability is always a top network priority, with some carriers promising a 99.999% uptime. The requirement for high availability and minimal network downtime imposes very strict requirements on switch reliability. Planar lightwave circuit (PLC) switches, which do not use any moving parts or chemical state transitions, are considered to be most reliable. To minimize network downtime during system reconfiguration, strictly nonblocking switches are being used. These switches do not reroute existing paths to establish new connections, thus eliminating any traffic disruption due to system configuration.

The use of smart optical switches can greatly speed and simplify network administration. Smart software can perform such tasks as dynamically adding and dropping customers to and from the network, reconfiguring the network, monitoring network status for actual or imminent failures, and adjusting the light intensity throughout the network.

Optical signals, entering the amplifiers present throughout the network, must be equalized to prevent signal distortions. Consequently, switch outputs are normally equalized by the use of variable optical attenuators (VOAs). The need for these external devices increases the total system cost, overall losses, space, and power consumption. The need for VOAs also reduces system reliability. Switches that support internal signal attenuation and dynamic equalization offer large cost savings.

Abe Queller is the vice president of customer applications and support for Lynx Photonic Networks (Calabasas Hills, CA). He can be reached via the company's Website,