Redirecting the light with intelligent optical switches
Switching software that can manage millions of light paths in a mesh architecture is a key challenge of intelligent optical networks.
RICK THOMPSON, Sycamore Networks
What's in a name?" asked an article that ran in Lightwave's supplement, Fiber Exchange (March 25, 2000, p. 13) some months ago, as the author explored the industry confusion regarding the differences between optical crossconnects (OXCs) and optical switches, terms frequently used interchangeably. Vendors, carriers, and analysts all weighed in on their preferred terms for these devices, which redirect light from optical input and output points.
The tentative consensus was that the term "crossconnect" generally depicts the old-fashioned or first-generation patch panels, while "switch" refers to the newer technology, which offers routing and management functions and is designed for use in a mesh architecture. The line between the two terms is not completely drawn, however.
What's clear is that the debate over these terms is arising from a change in communications networks, as crossconnect/switching technologies evolve to relieve the strains of an overloaded public network. Defining the differences between OXCs and optical switches will help to explain where these new technologies fit in next-generation networks.
The new crossconnect/switch technologies will be used for different applications. Some of these devices will serve as digital-crossconnect replacements, which assumes the network will continue to be built as it has been in the past. Others will be used as "intelligent" optical switches to address the new challenges of the optical core.
This year is pivotal for optical switching. Intelligent optical switching has moved from a concept on PowerPoint slides to products and now early deployment. The major challenge ahead is to develop switching software that can control and manage thousands of switches and millions of light paths as a single network that provides continuous availability.
In this application, service providers replace existing digital-crossconnect systems (DCSs) with optical crossconnects to upgrade and extend traditional SONET ring-based networks. The value of the new systems in this application is in reducing space and power requirements and providing cost savings over traditional crossconnect technology. This application does not require software intelligence; the goal is to simply groom low-speed services onto a high-speed backbone in a static network environment.
These new devices can extend the life of SONET/SDH rings, but will not help carriers evolve their infrastructures toward service-ready network architectures, since this equipment lacks networking intelligence and end-to-end point-and-click provisioning capabilities. While this strategy may offer an interim solution for carriers that have invested heavily in SONET/SDH ring architectures, it will eventually become untenable, as there are serious limitations to the SONET/SDH architecture with respect to efficiency and timely provisioning of revenue-generating services.
SONET/SDH ring architecture works well as the voice infrastructure it was designed for with its strong failure-protection mechanisms, namely the concept of rings. But that essentially cuts in half the available bandwidth for working traffic. In the SONET/SDH model, there must be protection bandwidth equal to the bandwidth for working traffic available on the same ring. With organizations pushing the need for multiple OC-3/STM-1 (155 Mbits/sec) and OC-12/STM-4 (622 Mbits/sec) at the access of the network, protection bandwidth quickly becomes a limiting factor in the available capacity of a carrier network.
In addition, when traffic increases on a SONET/SDH network to meet these demands, every ring in the network must be upgraded, resulting in a costly and time-consuming process. As a result, there is no mechanism for automatically provisioning and managing a circuit from end-to-end in the ring architecture. The connection must be manually mapped at each point where one ring is connected to another. Management frames must also be mapped from one ring to another. This task is labor-intensive and needs to be done carefully to make sure the proper connections are made. Because digital- crossconnect replacements maintain the SONET/SDH architecture, these devices do nothing to alleviate these architectural problems.
The industry is moving away from SONET/SDH rings and toward mesh architectures to solve these problems. Digital-crossconnect replacements, however, are not designed to migrate to mesh networks. These devices provide some beneficial functions such as reducing size, power, cost, and prolonging the existing network architecture. But they do not provide a migration path to the new mesh architecture the industry is moving toward.
The intelligent optical switch, on the other hand, solves both problems by serving as transition equipment in legacy infrastructures and offering true mesh capabilities. Generally, intelligent optical-switch applications focus on the migration from a static environment to a flexible mesh architecture. In these applications, service providers want to deliver dynamic bandwidth services ranging in speeds from STS-1 to OC-192 (10 Gbits/sec) today and moving to higher data rates in the future. An Ethernet hierarchy must also be supported. The word "intelligent" before the optical switch refers to the sophisticated routing and signaling software used to create and manage a flexible mesh infrastructure to deliver services in real time. Today's intelligent optical-switching applications utilize electro-optical switching products. Optical bandwidth is deployed and re-deployed in a moment's notice to meet changing customer demand. In effect, creating "fluid bandwidth" and delivering true services on demand.
As all-optical switching technology matures and it becomes possible to extend software control into an all-optical environment, higher-speed services will migrate to an all-optical architecture to achieve greater cost savings and scale. First-generation all-optical switches, in the concept stage, attempt to reduce costs in provisioning light-based services but have limited value because most do not address the software intelligence issues.
The next-generation of intelligent switches add routing and management software to automate provisioning, routing, and restoration of light paths, along with integrated SONET/SDH functions. That enables any-to-any node connectivity and multilevel, nested protection schemes on a port-by-port basis. The goal is to provide the most efficient use of wavelengths and fiber through subrate grooming, Express Trunk association (efficiently packing subrate services onto wavelengths), and network optimization advice.
Next-generation intelligent optical switches will offer the following characteristics:
- Scalability. The switches are built to grow as the data traffic increases.
- Nonblocking architecture. For every available service input, there is an output path, which does not disturb existing services along the way.
- Interconnection in hybrid environments. These switches can communicate network control information either in-band or out-of-band.
- Intelligent routing. These devices have intelligent, IP-centric, standards-based routing and signaling algorithms embedded to optimize light-path placement.
Used in intelligent optical networks, intelligent optical switches provide data-like networking in the optical domain and traffic engineering at the wave level, which allows carriers to optimize light paths. The result is scalable networks that can expand as IP traffic grows and offer better performance due to the minimization of hot spots and failure points.
It is important to understand that scalability has more than just one dimension. Hard optics defines the node scalability, or the matrix size of a single switch. Soft optics defines the network scalability, or the size of the entire global network.
Service providers are building large networks composed of many elements of varying sizes. The larger the network becomes, the more intelligence it requires for smooth operation.
Each of the network elements must be aware of the other elements' bandwidth characteristics and how that bandwidth interconnects with the network's switches. This is the only way services can be delivered end-to-end over an intelligent optical network.
However, the transition to these optical mesh networks will not happen overnight. The Table offers a transition timeline based on the emergence and commercial availability of next-generation optical-switching solutions. Many networks must evolve from pure SONET/ SDH to fully meshed architectures. Service providers have invested heavily in current platforms and need a migration path that allows the use of existing infrastructure, while laying the groundwork for the mesh architecture. During this migration, the switching devices will be used to connect the underlying optical transport network to IP routing devices.
Ultimately, intelligent optical switches will serve as critical elements in enabling end-to-end optical-networking solutions, which satisfy the following next-generation architectural requirements:
- Multiservice. Future networks must be capable of transporting and switching SONET/SDH as well as Ethernet (1 Mbit/sec to 10 Gbits/sec), Escon, and a variety of other signals.
- Robust and futureproof. Next-generation architectures must seamlessly adapt to new environments and technologies and the use of the "right" technologies for different parts of the network. Modular, scalable hard-optic designs in combination with standards-based soft optics will enable component innovation to be quickly introduced into value-added systems. These technologies will also allow service-enabling software features to be added to intelligent optical networks, as needed.
- Efficient. For cost-effectiveness, wavelengths need to be used efficiently. For instance, if a wavelength is capable of supporting 40 Gbits/sec, it should not be wasted carrying a 2.5-Gbit/sec signal. The mesh routing and restoration algorithms also need to make efficient use of network resources by intelligently packing subrate lambdas originating at the edge over high-speed light-path services in the core.
- Service-level granular switching. The network architecture must seamlessly support different service rates over all portions of the network. The hard optics in certain network elements will define a minimum service granularity, which may differ from that of other equipment. A common network-wide soft-optic platform is needed to ensure end-to-end service provisioning in such network architectures.
- Protection and restoration. The network should provide both dedicated automatic protection switching and efficient mesh level restoration, with various levels of protection in between. It is critical to create tiered levels of protection and restoration for carriers to offer different classes of services over intelligent optical infrastructures. That will enable next-generation networks to offer a platform for priority-based services, opening new opportunities to turn bandwidth into revenue and to bill accordingly.
- Advanced networking services. As networks evolve, advanced services such as optical virtual private networks and customer network management will allow carriers to differentiate services. Today, these advanced networking services are available on software-centric platforms.
- Carrier class. The network should provide carrier-class performance, reliability, scalability, and management. Other requirements include network element auto-discovery, point-and-click provisioning, high availability, reliability, robustness, detailed performance monitoring, and rich diagnostics as well as advanced fault-isolation features. Too often, carrier class is looked upon as synonymous with fault-tolerant hard-optics. Carrier class is taking on new meaning with soft-optic architectures that allow software modules to be started and stopped independently, upgraded on the fly, and segmented to pinpoint and fix potential network problems without bringing down existing light paths. This is another area where digital-crossconnect replacements differ from intelligent optical switches. While digital-crossconnect replacements solve the hard-optics issues related to carrier-class gear, intelligent optical switches take networks to the next level by attacking the carrier-class software issues.
As networks grow in size, more intelligence is required. To achieve scalability, efficiency, and increase network intelligence, switches will permeate next-generation networks. As mesh architectures expand, a large global network could entail large switches in several hundred core cities and tens of smaller switches within each city and its suburbs. Switch size and service granularity will define the hard-optics requirements of these various-sized switches.
The behind-the-scenes key to the success of these switched networks is the software. The ability to reach the end points of the network, under a common soft-optic architecture, will provide a scalable environment, allowing service providers to differentiate advanced revenue-generating services.
Rick Thompson is the director of core switching and management at Sycamore Networks (Chelmsford, MA).