Third-generation DWDM networks near reality

March 1, 2001

The next wave of DWDM network equipment will provide fast and flexible service provisioning and reduce costs.

DR. DOUGLAS J. ARENT, Network Photonics,
and ARLON MARTIN, Agility Communications

While bandwidth growth and network capacity continue to accelerate in the LAN and WAN, challenges remain in the metropolitan area. Here, SONET equipment and new "next generation" or second-generation DWDM gear struggle to address service-provider needs and offer the capabilities to handle applications requiring high (gigabit)-bandwidth, flexible network configurations. Third-generation networks, using dynamically configurable and manageable all-optical DWDM innovations, offer service providers new tools to address this pressing opportunity.

Speed refers not only to bandwidth, but also to speed of service. Both are of increasing importance. Customers want and will pay for network services that deliver high-bandwidth capacity when and where they need it. What remains to be seen is who can deliver those services and when.

The first service provider to deliver high-bandwidth capacity will capture significant first-mover advantages, build customer loyalty, generate new revenue, and open opportunities for additional enhanced services. To meet customer demands, service providers must look for new capabilities and address increasing capacity bottlenecks, particularly in the metropolitan area.
Figure 1. The first generation of DWDM is characterized by point-to-point links targeted at relieving fiber exhaustion.

Bandwidth has continued to grow in the LAN-from 10/100Base-T to Gigabit Ethernet and soon 10-Gigabit Ethernet-and in the long-haul backbone-from simple SONET services to high-capacity DWDM systems that now support terabit transfer rates. However, customers still struggle to connect geographically separated LANs or obtain high-speed connectivity to the backbone through the metropolitan-area network (MAN) due to its limited capacity (e.g., megabits to a few gigabits per second).

Service velocity-the speed at which service is provisioned or changed-today is best described as pitiful. Customers want multimegabit service but instead must wait months for simple T1 circuits. DS-3 and OC-3 through OC-12 service wait times approach six months in many locations. To make matters even worse, traffic patterns are changing, further increasing the strain on metro networks. With increased Internet/data traffic, caching, and the introduction of application service providers (ASPs), storage-area networks (SANs), and other services, metro traffic growth will continue to explode.

Exploding bandwidth has quickly made obsolete traditional SONET networks that support only single-wavelength ring architectures. SONET grooming and framing that worked well for TDM traffic has proven increasingly inefficient for multimegabit and gigabit data traffic. Moreover, metro SONET architectures were designed primarily to transfer local traffic to long-haul networks, concentrating inter-ring switching at the long-haul crossconnect. While second-generation DWDM metro networks begin to address bandwidth requirements on a given ring and incorporate static optical add/drop multiplexers (OADMs) to allow wavelength-based point-to-point services, today's emerging networks are driving the need for large-port-count (e.g., 1,000x1,000 or larger) all-optical crossconnects (OXCs) to switch traffic from ring to ring. That is at best an interim and costly solution.

Enabling the network with distributed switching capabilities and supporting ring or logical mesh architectures significantly reduces the requirements for the OXC and allows networks to be optimized for increasingly variable traffic patterns. Fiber is no longer dedicated to backhaul traffic, and port-count requirements at primary switching hubs are significantly reduced.
Figure 2. The second generation of DWDM, now reaching the field, adds support for protected ring architectures and a variety of service interfaces to the ability to tackle fiber exhaustion. Optical add/drop capabilities have also arrived. This equipment at least partially addresses metro-network concerns.

Networks that offer dynamic, all-optical wavelength management will enjoy significant advantages over next-generation systems. Dynamic management brings to fruition full add/drop capabilities, protection and restoration, wavelength routing, and performance monitoring. All services are available on a per-wavelength basis. Service providers employing third-generation systems based on all-optical dynamic wavelength management will be able to offer an order of magnitude more bandwidth per fiber, dramatically improved service velocity, optical-layer protection, and wavelength-level services, which all lead to significant economic advantages.

As depicted in Figure 1, first-generation DWDM equipment has been the technology of choice for long-haul networks. These relatively straightforward point-to-point systems have rapidly expanded from a few wavelengths at OC-48 to upwards of 100 wavelengths at OC-192, offering capacities as high as 96 OC-192 wavelengths (960 Gbits/sec per fiber). Driven by capacity constraints and the high cost of laying new fiber, service providers installed point-to-point DWDM gear to optimize the capacity per fiber-mile. This approach helped solve the fiber-exhaust problem, but it also added an additional equipment layer and additional cost to the network.

Second-generation DWDM equipment has recently emerged to offer the capacity benefits of DWDM to MAN service providers. Adding low-wavelength-count second-generation DWDM to the MAN helps relieve fiber exhaust, similar to long-haul fiber relief, as shown in Figure 2. More important, network architectures employing second-generation DWDM capabilities now support protected ring architectures and offer multiple service interfaces such as Gigabit Ethernet, Escon, Ficon, Fibre Channel, and OC-3 to OC-48.

While these improvements seem monumental compared to SONET networks, even second-generation capacities (e.g., 80 Gbits/sec per fiber pair) are expected to last only 12-18 months. Three additional problems persist with second-generation DWDM networks: cost, scalability, and manageability.

Three expenses dominate most service providers' budgets: fiber, equipment, and OAM&P (operations, administration, maintenance, and provisioning). Operationally, service providers spend between 40% and 80% of their total network costs on OAM&P due to the intensively manual nature of first- and second-generation network equipment. For example, interface cards with static optics that are unique for each deployed wavelength require truck rolls and card removal/replacement to reconfigure wavelength routes. Additionally, connecting new circuits typically involves manual patch-panel configuration and extensive network operations center updates and can take up to six months.
Figure 3. The third generation of DWDM enables crossconnected ring and mesh architectures using dynamic wavelength crossconnects and optical add/drop multiplexers.

To address this challenge, service providers are requiring end-to-end management systems that support remote provisioning, activation, and fault correction. Inventory expenses are also considerable, owing to the fact that lasers and filters are static, forcing service providers to keep considerable amounts of inventory on hand-40 spare transponders for 40 wavelengths can be required per node. Other issues include rack-space rental charges and power, both of which can be deal breakers when securing new collocation space.

While costs affect the bottom line, top-line revenue growth is limited by scalability and manageability. Without available bandwidth, service providers cannot respond to new service requests, and coming in second is equivalent to losing the race. Likewise, service providers with low service velocity are at a competitive disadvantage to companies that offer timely connectivity.

One of the principle bottlenecks in second-generation networks is the OXC. Conventional wisdom-and the focus of considerable R&D and investment-is driving the continued production of large, dynamic-switching OXCs. Most OXCs today use an optical-electrical-optical (OEO) approach due to the lack of standards for optical interconnection of wavelengths in multivendor environments. OEO switches offer limited scalability and tend to be expensive solutions due to the requirement for demultiplexing the DWDM light streams, converting them to electrical signals, switching, reconversion to the optical domain, and remultiplexing.

As network capacities continue to explode from megabits to hundreds of gigabits to terabits, only all-optical switches offer the required capacities and investment protection. Network architecture advancements can reduce this dependence on the OXC for switching operations in the MAN.

Third-generation DWDM networks provide scalable, all-optical, distributed wavelength switching. Fundamentally different from second-generation architecture, third-generation DWDM architecture specifically addresses costs, scalability, and manageability. Moreover, third-generation networks enable service providers to maximize revenue through new, differentiated services while simultaneously reducing operating costs. Within metro and regional areas, interconnected rings and logical meshes may be optimized by using all-optical wavelength crossconnect switches (WXCs), dynamically configured OADMs, tunable-laser transponders, and advanced management software to control the dynamic optical layer.

Service providers are able to dynamically route any wavelength to any node on any network link, transforming DWDM from a fiber-exhaust solution to a system of commerce. Wavelength-agile third-generation networks will allow the ability to manage wavelength routes among interconnected rings or through mesh architectures while eliminating the need for centralized switching at a large OXC.

As shown in Figure 3, local traffic (λ8) remains in the MAN, and optically traverses multiple network elements. Long-haul-bound traffic (λ15) connects to the long-haul network through the OXC. This wavelength-level manageability has three effects: reducing the need to connect all wavelengths to the OXC, thereby reducing port requirements for the OXC; enabling instantaneous service velocity through remote provisioning and wavelength tunability; and lowering OAM&P costs due to the availability of universal transponder cards (with tunable lasers and filters) and remote provisioning.

Another benefit of third-generation DWDM systems is the ability to optimize the reuse of wavelengths. For example, in Figure 3, λ3 is active on a link on access ring 1, inactive in the core, and active again on access ring 3. Network protection can utilize both switching and tuning to route around failed nodes or links and as an enhanced service option is available on a per-wavelength level.

Dynamic routing supports many emerging services such as streaming video, SANs, and optical virtual private networks. For example, an ASP may offer data storage and disaster-recovery services to a set of customers within a metropolitan area (see Figure 4). Each customer only requires a temporary connection to the ASP while the data transfers occur. The ASP can switch its connections from Company A to Company B, while only using a single wavelength (λ6). Emerging signaling protocols such as Multiprotocol Lambda Switching will make such a scenario very possible.
Figure 4. In this example, an ASP can use the same wavelength to serve two customers. Only third-generation DWDM equipment makes this strategy economically attractive.

Third-generation networks also bring advances in the optical domain, including optical protection of individual wavelengths and optical performance monitoring, through the use of forward error correction (FEC) and digital-wrapper technologies. Optical protection capabilities address wavelength level, path, or link level, and allow service providers to differentiate and guarantee service levels. Similarly, FEC and digital-wrapper technology bring two significant benefits to third-generation service providers: (1) performance monitoring, such as bit-error rate and path-link status for enhanced, verifiable service-level agreements (SLAs), and (2) bit-rate- and protocol-independent wavelength services.

SLA offers are further enabled through the ability of third-generation networks to provision services extremely quickly, such as through point-and-click provisioning and route setup on the order of minutes. Service differentiation granularity may be defined through wavelength-level protection, including switching, path protection, and wavelength redundancy (tuning) around failed nodes or links. Each of these advances, combined with fundamental architectural developments, enable service providers using third-generation networks to offer and deliver a wide variety of differentiated services while also dramatically improving overall profitability.

Third-generation networks bring to bear three significant technical achievements: dynamic wavelength switching, tunable lasers, and tunable filters. Wavelength-switching nodes provide the critical enabling feature of provisioning any wavelength to any node using completely and dynamically reconfigurable OADMs and WXCs. Working in the optical domain, with advanced power-balancing algorithms to ensure that optical power budgets are met, wavelength-based services may be provisioned in real-time via software activation.

These wavelength-based services offer protocol and bit-rate independence supporting Gigabit Ethernet, 10-Gigabit Ethernet, OC-3 to OC-192, STM-1 to STM-64, Escon, Ficon, and Fibre Channel, enabling the provisioning of any wavelength to any node throughout a complex network of interconnected rings or meshes-without the need to convert from optics to electronics or to be switched through a central OXC.

The second significant technical advancement is tunable lasers. Optimally, tunable lasers offer a broad range of tunability over all wavelengths using a single laser, high enough transmit power to minimize the need for erbium-doped fiber amplifiers, and fast tuning speed.

Fiber-optic networks have historically been static and could only accommodate long-distance networks using lasers fixed on a given channel-with no flexibility to dynamically provision traffic to other channels. With fixed-wavelength lasers, service providers had to know which channel was being used and never change it, regardless of the rest of the traffic on the network. When service providers needed to change the fixed laser for any reason, they had to pull the old board out and install a new board-always having spares available just in case. The growing size of optical networks has made this old method cumbersome and increasingly inflexible.

After the industry learned how to put multiple wavelengths on a laser, the ability to support faster increases in network capacity opened up. Narrowly tunable lasers that can tune across 20 to 30 channels are being deployed to reduce the complexity of planning, control inventory, and better accommodate changing network demands.

Yet, as bandwidth demands and network traffic continue to increase and change rapidly, narrowly tunable lasers have also become increasingly inefficient. The space requirements, complexity, and cost of using lasers tunable over only 20-30 channels also makes change and growth of large networks difficult and costly. The optical-networking industry now requires more powerful laser technology that can support long distances and far greater range in wavelength than are currently available from narrowly tunable lasers.

Widely tunable lasers based on sampled grating distributed Bragg reflector (SGDBR) laser technology have come of age and can deliver much higher power than is generally available in the market today to any one of 100 channels. These breakthrough lasers enable service providers to:

  • Rapidly provision wavelength and bandwidth with the power to address a wider range of applications.
  • Significantly increase network capacity by supporting the entire C-band from 1,525 nm to 1,565 nm with 50-GHz spacing, enabling the user to select one of 100 available channels dynamically.
  • Reduce the size and cost of replacement transponder inventories by enabling a single transponder to support all channels rather than requiring unique lasers to support each channel.
  • Streamline the planning process for network growth by enabling communications companies to add any channel to the network infrastructure, knowing it can be used to enhance capacity through dynamic switching.

Tunable filters are the third optical technology advancement required in third-generation optical networks. Tunable lasers enable the transmitter to pick an available wavelength, and wavelength switches allow the connection to be made between the transmitter and receiver. Tunable filters allow the receiver to tune to the required wavelength to complete the link. A variety of tunable-filter technologies are emerging such as tunable fiber Bragg gratings, tunable-cavity filters, and tunable thin-film filters.

Third-generation networks, using dynamically managed, all-optical DWDM innovations, offer the ability to deliver differentiated services based on wavelength-level granularity. Combining advancements in dynamically manageable OADMs, WXCs, tunable lasers and tunable filters, optical performance monitoring, wavelength-level protection options, and network-management software, third-generation networks enable service providers to address the need for speed. Service providers may now offer new revenue-generating services while cutting operating costs.

Douglas J. Arent is director of strategic marketing at Network Photonics (Boulder, CO), and Arlon Martin is vice president of marketing at Agility Communications (Santa Barbara, CA). They can be reached at their respective e-mails, [email protected] and [email protected].