Bandwidth tornado, .com tsunami, Moore's law of data rates--all of these descriptions fail to do justice to the growth of the Internet and its bandwidth--hungry applications, a phenomena that is pushing the current network infrastructure to its limits. Long-haul networks were first to suffer the effects of an over-burdened infrastructure. The solutions offered by startups as well as established vendors--point-to-point dense wavelength-division multiplexing (DWDM) and bandwidth managers--have found widespread acceptance among interexchange carriers (IXCs). To avoid a similar capacity problem in the metropolitan segments of these infrastructures, network planners need to adopt flexible, scalable solutions that can accommodate evolving user and application requirements.
The metropolitan network connects to a long-haul infrastructure (part of a transcontinental backbone) on one end and to the users on the other end (see Fig. 1), which fundamentally alters the requirements placed on metropolitan area networks (MANs). In addition to supporting diverse users and their applications, the geographical boundaries of the metropolitan infrastructure differ from those of long-haul networks and thus require different technologies for capacity and connectivity. To provide connectivity between MANs, the primary requirement placed on the long-haul infrastructure is to offer capacity at the lowest cost-per-bit possible and, of course, a way to manage this capacity.
The generally accepted definition of the metropolitan network divides it into two segments--the metropolitan core and the metropolitan access (see Fig. 2). The core network covers areas up to a few hundred kilometers in circumference, which is sufficient to provide connectivity in large urban areas. Metropolitan access networks consist of point-to-point, ring, or meshed subnetworks that connect the various users in the metropolitan areas. The access portion uses fiber as the primary means of transmission.
Given the diverse range of equipment and services that need connectivity in the metropolitan domain, it's critical to know what user and application requirements the network must fulfill. The metropolitan network connects to many types of end-customers: consumers, enterprises, corporate campuses, and Internet service providers (ISPs), among others. These users/applications generally fall into the following segments: consumers, business-to-Internet, business-to-business, ISPs, and interoffice (see Fig. 3). Each of these segments has specific network requirements.
While the last-mile connections to the home remain mostly copper, new technologies such as digital subscriber line are quickly multiplying the magnitude of data bandwidth originating or terminating at each consumer's home. As these circuits are aggregated, the required bandwidth quickly scales to a level (>155 Mbits/sec) where fiber is the best transmission medium. Even without the fiber-to-the-home and fiber-to-the-curb fanfare in recent years, fiber access is moving closer to consumers.
The targets of multibillion-dollar business-to-consumer e-commerce, home users need cheaper and faster access to the Internet and other online services. As a result, the requirements placed on the metropolitan network are twofold: low first-cost and scalability. Low first-cost makes high-speed data connectivity solutions economically feasible for the carrier. Scalability allows rapid facilities growth without requiring new builds of equipment.
Today, most successful businesses require broadband connectivity to the wide area network. The growing segment of business-to-business e-commerce is further fueling the need for bandwidth and intelligent connectivity. The business-user segment places the most critical requirements on the metropolitan infrastructure because it includes the paying early adopters of new technology and services, who ultimately drive growth.
Business-to-Internet and business-to-business users require the following characteristics in metropolitan networks: manageability, data-centric, service transparency, dynamic provisioning, and survivability. While manageability is an easy characteristic to understand, it is the most overlooked at the infrastructure level. Claims of hardware excellence from vendors are rarely backed up with network operating systems that can provide end-to-end route managability in the metropolitan area. Business users require a metropolitan network that is engineered to carry Internet-protocol (IP) traffic, and is data-centric. IP traffic is the primary currency for e-commerce and the type of data that dominates bandwidth as well as revenue charts. Service transparency is also key for metropolitan networks so that any interface can carry either Synchronous Optical Network/Synchronous Digital Hierarchy (SONET/SDH) or Gigabit Ethernet (GbE) traffic. This transparency allows the network, for example, to support GbE links between corporate local area networks (LANs) as well as Fibre Channel links to server farms.
Dynamic provisioning of capacity is another fundamental requirement for metropolitan networks. The metropolitan environment must adapt readily to constant change as the number of business users and business locations grows or fluctuates. The network should be flexible enough to accommodate bandwidth deployment on an "as needed, where needed, when needed" basis. For example, a company's multiple locations may need high bandwidth during business hours for internetworking and more capacity on other routes during the night for corporate data/server backup.
Finally, because mission-critical applications run over metropolitan networks, survivability is another key factor. Specifically, the restoration of services in less than 50 msec-a standard set by sonet/sdh rings-is a basic requirement.
In a healthy competitive environment, multiple service providers flourish in order to provide the best services to end users at the right price. As large service providers begin to carry significant amounts of IP data and as voice, video, and data are delivered over this unified pipe to businesses and consumers, efficient transmission requires a metropolitan-optical-network infrastructure. The specific requirements range from scalability for handling the rapid growth of data-centric traffic to manageability for supporting the efficient introduction of new services.
Traditionally, metropolitan networks have provided connectivity between the central offices of incumbent carriers. As fiber routes with sonet/sdh rings begin to reach their limits, additional capacity is needed to provide fiber relief, scalability, and in-service increases of bandwidth.
A successful metropolitan solution must address all of these requirements to be relevant to incumbent and competitive carriers in the metropolitan arena. Carriers who meet these criteria are more likely to avoid inefficient and costly infrastructures.
It is useful to note that these requirements have more in common with the flexible and dynamic nature of a LAN than with a long-haul network. Most of these requirements are inherent in LAN systems. Only the capacity and scalability requirements are met in traditional telecommunications and long-haul networks (see Fig. 4).
The common theme underlying metropolitan network requirements is "flexibility." Unlike a long-haul network solution where each link is tailored to operate at a specific bit-rate and distance within a fixed topology, metropolitan optical links must accommodate all possible bit rates as well as topology changes without requiring a change in equipment. As competitive carriers seek to minimize the number of truck-rolls per customer order, software-provisioned bandwidth becomes a necessity along with multiservice transport ports capable of handling any bit rate from OC-3 (155 Mbits/sec) to OC-48 (2.5 Gbits/sec), including data-optimized interfaces such as GbE.
Rapid restoration in the event of fiber failure requires flexible optical-ring solutions. This flexibility not only provides fast survivability, it also enables scalable services and in-service upgrades, while adding either traffic at a node or more nodes to the network. Metropolitan optical rings should support a diverse mix of ring-protection switching topologies such as unidirectional-path and bidirectional-line switching. To support existing SONET/SDH rings over optical rings, the carrier has a choice: disable the optical-layer protection or have the optical rings seamlessly interoperate with the existing SONET/SDH infrastructure.
For scalability and capacity at the lowest possible initial cost, DWDM techniques can increase the number of channels supported per node. Band-organized DWDM, where a small number of channels can be added or dropped per node, is a useful example of this type of technology. While other technologies such as subcarrier multiplexing can also provide higher bandwidth, DWDM remains the proven solution for telecommunications networks with equipment from multiple vendors.
In addition to the advanced optical transport hardware, the metropolitan infrastructure needs sophisticated network operating systems to manage the dynamic deployment of services and bandwidth as well as the real-time performance and the reliability expected of carriers' telecommunications systems. Somewhat analogous to a personal computer that has a fixed set of resources allocated according to the demands placed on the system via the keyboard (user) or applications, the metropolitan network's dynamic scenario calls for a similarly agile and fast network operating system. This system arbitrates between various users and other network demands according to the policies set by the carrier, helps the carrier manage end-to-end services over the metropolitan infrastructure, effectively unites the infrastructure and service management aspects of a telecommunications system, and provides open interfaces to the carrier's fault-management, billing, and tariff systems.
The final part of the metropolitan solution is an information link between the data routing and switching layers and the transport layer. A clear separation of responsibilities between these layers is necessary for the carrier to avoid replicating complex packet routing and flow-control decisions made in the data routing/switching layers. But for the data layers to operate intelligently, they need to be logically coupled together and linked to the physical optical layer where information about network topologies, channel deployment, fiber or facility failures, and protection mechanisms resides.
The combination of a next-generation transport system, an optical network operating system, and internetworking protocols is critical for meeting the requirements of the MAN user. Without equal emphasis on each of these three elements, the networking solution is incomplete and unable to deliver the desired set of services at the right price to consumers and business customers in the metropolitan area.
Rohit Sharma is vice president and founding chief architect at Optical Networks (San Jose, CA).