IP ATM & SONET/SDH
As carriers rapidly emerge, the wide array of network solutions indicates roles for SONET/SDH, ATM, and IP.
In economies where nearly every facet of production, delivery, and consumption is shaped by telecommunications, the opportunity for new carriers to emerge in competitive markets is almost limitless. Faced with competing against entrenched incumbents, new carriers demand a lot from equipment vendors: a low initial systems cost, mechanisms for rapid delivery of services, flexible allocation of bandwidth, varying levels of restoration and protection, and scalability to accommodate Internet-scale growth.
Current Synchronous Optical Network/Synchronous Digital Hierarchy (SONET/SDH) and Asynchronous Transfer Mode (ATM) gear, however, often discourages these carriers on all fronts. This situation has encouraged a plethora of optical-networking start ups to create unique core, edge, and access systems that ideally offer all these capabilities by leveraging improvements in system integration, optical components, and networking protocols. Although convergence is finally more a reality than a buzzword, the optical-networking market has never been more diverse.
This diversity begs an important question: With so many new carriers striving to be unique in a highly competitive market, is it possible, or reasonable, to assume there is an ultimate solution for optical networks?
The experience of vendors in the past three years is demonstrating that pronouncements of the death of SONET and ATM, in the wake of Internet protocol (IP) and transparent optical networking, is simultaneously true and false. SONET is considered far too inflexible to accommodate the dynamic bandwidth provisioning and allocation required of new carriers' networks. Nevertheless, SONET recorded extraordinary growth in 1999, driven by equipment sales to carriers that favor its standardization and reliability. Similarly, ATM, once considered the unifying switching and transport vehicle for the entire public network, is finding new application in integrated access devices. At the same time, the cell-based protocol is experiencing a diminishing role in many core networks, forced out by the packet-based IP and its promise of a unified network.
In 1996, the optical-networking market consisted solely of dense wavelength-division multiplexing (DWDM) systems deployed by a handful of long-distance companies. Today, there are at least six distinct market segments-the core, backbone, metro/regional, edge, access, and enterprise-and within each, carriers have more choices than many in the industry could have imagined just four years ago (see Table 1). The core includes those network elements that provide the switching and routing of optical-layer circuits. These systems tend to be optimized for a mesh network architecture and act like broadband digital crossconnects, only the circuits are primarily switched at rates equal to, or above, an OC-48 (2.5 Gbits/sec). The backbone market consists of multichannel long-haul transport, the driving force of the optical-network market to date. The metro/ regional segment consists of systems interconnecting central offices, Internet service provider (ISP) points-of-presence (PoPs), colocation centers, and hubs in a metropolitan area.
The optical edge is a newly emerging category, which includes next-generation metro equipment. This equipment integrates many disparate network elements into a single device. Typically, network traffic is carried through the metro area on WDM channels. Optical access networks support individual clients, buildings, or access nodes over an optical network, providing a highly flexible and scalable infrastructure that brings broadband services to small and medium-sized businesses on par with today's major corporate customers. This low-cost equipment includes wavelength-division multiplexing (WDM) systems and optical amplifiers.
Enterprise optical networks devote individual wavelength channels to such protocols as Fibre Channel, enterprise systems connection (Escon), and Gigabit Ethernet (1 Gbit/sec) for the interconnection of corporate data centers or storage area networks. Equipment in this segment includes low-cost WDM and technologies that help map Fibre Channel, Escon, and Gigabit Ethernet to optical channels.
Over time, more segments may emerge, particularly on the access side. As optical-layer technology becomes more affordable, it is easy to foresee new products that create optical-layer, campus, and local area networks (LANs), devoting a single wavelength to individual workstations, private branch exchanges (PBXs), or storage arrays. In the residential market, technologies are already being developed that devote a single wavelength to curbside optical-network units, or in some cases, individual dwellings.
As carriers evolve their networks to better accommodate the transition of telecommunications traffic from circuit-switched to packet, the opportunity for an optical layer to perform both transport and switching functions becomes more attractive.
Today, aggregated signals must be processed entirely at network nodes, regardless of how the traffic is handled, adding expense, complexity, and potential for failure. The goal of the new public network is to create a transparent optical layer that can easily manage the transport of signals entirely in the optical domain, reducing the need for processing and interpretation of signals, while increasing the reliability and restorability of high-bandwidth networks.
What remains a highly contentious issue in the networking industry is how existing technologies and network protocols will exist within this new network architecture. There are currently three different perspectives regarding optical networking. One is the prevailing SONET/SDH-centric model, in which SONET/SDH is the unifying transport layer for circuit, ATM, frame relay, and pure IP services. This model is at work in the vast majority of carrier networks, in which overlay voice and data networks are unified at the core by SONET/SDH.
The second perspective is an ATM-centric model, in which ATM provides the core bandwidth and traffic management as well as the switching and routing functionality in distributed networks. This situation is most commonly realized in large Internet backbones, where IP routers exist at the edge and ATM switches are at the core. These "routed-edge, switched-core" networks are sufficient for today's applications and bandwidths but do not scale well, creating opportunities for optical-layer routing and switching.
The third perspective, an IP-centric model, is often espoused by IP-centric vendors but has yet to gain the favor of carriers, particularly the incumbents, because of the immaturity of the IP standards. The IP-centric model is, however, very attractive to newer carriers seeking to achieve the lowest cost per bit while offering the widest variety of services over a common platform and architecture.
While new carriers often espouse a simple, unified network solution, these providers often make the picture more, not less, complicated. New carriers are propelled into this market by two powerful forces: widespread demand for high-speed connectivity beyond traditional large business customers and a host of venture-backed startup equipment vendors with nontraditional solutions, hungry for some customers. Each new vendor is compelled to offer an optimal solution designed to gain wide market acceptance, but that may quickly become an elusive goal. Particularly since the similarities between carriers are fewer than ever before, this fragmentation creates niche opportunities but at the same time limits the addressable market for new products. Table 2 indicates how the use of different protocols and DWDM varies widely among carriers.
Nowhere is the effect of this fragmentation more evident than in the emerging optical-edge market. The segment itself is difficult to define. At its simplest, the optical edge is the merging of optical-network equipment with edge routing and switching into a single element. Here, the value proposition is straightforward and has already caught the attention of the venture-capital community and carriers. By integrating the functionality of many central-office (CO) network elements into one device, the carrier benefits from a unified management system, reduced space and power requirements in the CO, and a flexible solution that can accommodate service upgrades, changes, and capacity growth.
At their best, these systems provide the carrier with a toolbox from which to provision services. Cards in a shelf may support Gigabit Ethernet as easily as OC-12c ATM (622 Mbits/sec) or traditional DS-3s (44.736 Mbits/sec). When capacity increases at a particular node, new cards or provisioned wavelengths are added only to the node-the entire network does not have to be upgraded. Moreover, these systems do not bank on the success of a particular protocol or service. Therefore, a carrier's bets are hedged with a variety of service interfaces, many of which are software definable.
If such systems become widely available in 2001, it seems logical that rapid displacement of existing equipment would occur along with a flurry of contracts for new builds. But again, the heterogeneity of carriers will prevent that from occurring as a tidal wave of change-expect slow erosion of the installed base of systems instead. The reasons are as many as the number of carriers, but the major factors can be grouped into two broad categories:
- Carriers with a large installed base of distinct voice and data networking equipment will be reluctant to quickly adopt integrated solutions. Regional Bell operating companies (RBOCs), in particular, have resisted integrated solutions because these systems do not conform to the traditional telecommunications carriers' rigid operations and standards requirements. Major interexchange carriers (IXCs) may also be slow to adopt integrated solutions because, like the Bells, these companies tend to operate separate organizations for transmission and services. Each organization makes independent decisions regarding equipment purchases, thus systems that combine transport and service creation can confound the purchasing process by requiring each department's approval. New vendors will need to convince one department and let people inside the company sway the other department toward a commitment.
- Many carriers have close relationships with existing vendors and will be reluctant to embrace a solution from a startup that purports to improve upon the entire network architecture. Most carriers believe that the network is utilizing best-of-breed equipment; therefore, a compelling migration strategy is key to success in the optical-edge market. In short, a new vendor must demonstrate that a product will not force a carrier to change its existing network but instead show how it will allow the provider to comfortably migrate toward a more efficient network over time.
All of which leaves the new carriers, and there are many. These carriers will by and large react favorably to new solutions as a way to gain a competitive advantage over incumbents. New carriers have small network-planning organizations and are able to make purchase decisions easily, which favors the startup companies. If new players bring products to market that ease the provisioning of services, simplify bandwidth management, and offer a platform that allows the carrier to undercut the competition in terms of price, then these vendors will not be ignored.
In the battles raging between ATM, IP, and SONET/SDH, it is a safer proposition today to design equipment that accommodates all three protocols. That can be accomplished in two ways: with either a completely "open" optical solution that incorporates universal optical interfaces or with an integrated platform that aggregates a multiplicity of services onto an optical backplane for transmission through the metro area to the wide area network.
Over the next five years, carriers will be asked to choose among an often baffling array of network equipment, and equipment choices will likely be interpreted as the triumph of one networking scheme over the others. If recent history is a guide, however, at the end of five years, there will only be more choices, more carriers, and more protocols, not fewer.
Scott Clavenna is principal analyst at Pioneer Consulting LLC (Cambridge, MA). The company can be reached at www.pioneerconsulting.com.