The traffic-engineering and QoS capabilities offered by multiservice switch/routers, along with advances in the intelligent optical domain, promise a range of IP services.
Patricia Streilein and Joe John
Applications such as real-time multicast, virtual private networks (VPNs), e-commerce, streaming video, and wireless connectivity are gaining popularity. As a result, traffic patterns are changing and annual traffic growth is into triple-digit percentages. As networks continue to evolve at a rapid rate, the next marriage of technologies is clearly focused on providing flexible, simple, and rapid service delivery over a high-capacity infrastructure to support these applications.
Today, service providers can leverage connection-oriented data technologies such as ATM and frame relay to provide class-of-service differentiation and traffic engineering in the transport of these applications. But to build in support for these applications directly within Internet Protocol (IP), the protocol needs to move beyond "best effort" transport.
Technology advances are rapidly emerging in today's optical networks. Protocols such as differentiated services (DiffServ) for IP to support quality of service (QoS) and Multiprotocol Label Switching (MPLS) for traffic engineering are being developed. In the near future, a more intelligent optical domain will provide an infrastructure fed directly by multiservice switch/routers supporting Layer 2 services (e.g., ATM, frame relay, and point-to-point protocol) and Layer 3 services (IP) to direct traffic at the edge of the network.
Key to the evolution of next-generation networks is the interoperation of optical core and edge network elements to provide enhanced service delivery. That is accomplished by leveraging new technologies, such as dynamic optical signaling between data and optical devices, and Multiprotocol Lambda Switching (MPlS). Intelligence will provide optical crossconnects with the capability to maintain necessary QoS functionality through the optical core, as defined by the core or edge multiservice switch or IP router. In addition, further improvements for network operations are on the horizon. Advances in network-management systems are beginning to encompass data and optical devices, greatly simplifying provisioning.
As this network migration progresses, the multiservice switch/router is a key element in delivering services that support these applications. The multiservice switch/router automatically interacts with the intelligent optical core for higher-bandwidth transport. The optical core itself becomes a source of revenue by automatically providing high-bandwidth connection services for its clients.
In addition to low cost per bit, optical networking provides virtually unlimited bandwidth scalability to support increasing traffic demands. It also offers transparency for the multiple data protocols found in today's networking environment. Carriers that use numerous overlay networks for different traffic types can derive substantial benefits by migrating their networks to the more intelligent optical cores emerging today.
Early optical networks first emerged as fixed point-to-point networks built around DWDM systems. These networks were deployed to boost capacity in long-haul routes operated by interexchange carriers and backbone carriers. Later, as demand increased, DWDM-based optical networking found its way into regional networks.
Today, service providers are looking to bring optical networks even closer to end-customers-not only to increase capacity, but also to futureproof network architectures. As with long-haul and regional networks, metro optical networks will offer protocol and bit-rate independence, scalability, and rapid service provisioning.
With an increased number of wavelengths, carriers will spend considerable time and effort in provisioning and managing wavelength connections. While the newly introduced optical crossconnect systems will greatly reduce this effort, these devices are not enough. Today's unpredictable traffic patterns require network scalability at the core that only pure optical switches can provide.
With the advent of optical switches and software control solutions, today's optical infrastructures will evolve into intelligent networks. These networks will allow carriers to automatically provision services and manage traffic far more effectively. Intelligent optical networks will also provide cost-effective restoration mechanisms to support services with different QoS requirements under the operators' policy control.
Combining distributed intelligence, transport policy control, and lambda routing and signaling mechanisms will enable carriers to build optical networks with the operational flexibility of data networks. It will also allow carriers to manage higher-granularity traffic (e.g., 10 Gbits/sec) much more cost-effectively than traditional SONET networks. The evolving optical intelligence supports a number of key networking functions such as service provisioning, restoration, and performance monitoring. Carriers can enable new applications, including wavelength services, optical-layer internetworking, optical virtual private networks (OVPNs), and bandwidth trading and reselling to generate additional revenue. These new services are key drivers of the new intelligent optical layer.
Optical-layer internetworking allows high-capacity interconnections between large-core switch/routers across the wide area network. Much like MPLS, which provides traffic engineering with dynamic connections in the packet/cell-based network, MPlS will enable dynamic connections at the optical layer. This approach requires signaling and routing intelligence within the optical network to rapidly set up and take down lambda connections based on demand and traffic flow. For instance, a switch/router supporting dynamic optical signaling-standards for which are currently being specified by the Optical Domain Service Interconnect, Optical Internetworking Forum, and Internet Engineering Task Force-will be able to signal the optical network for additional bandwidth (i.e., more wavelengths) during periods of congestion.
The term VPN implies a logical rather than physical network. The intent of a VPN is to function much like a private network without the high cost of leased private lines. An OVPN operates across multiple managed wavelengths as a network that can be dynamically configured and taken down to support numerous applications. Unlike other forms of VPNs, an OVPN provides practically unlimited bandwidth across optical links using wavelengths. The end-customer has a VPN that can be completely protocol- and bit-rate-independent. Depending on available resources, an intelligent optical network can provision an OVPN in a matter of minutes compared to the months it would take to physically establish a SONET-based optical service connection.
Bandwidth trading and reselling is made possible by the deregulation of the telecommunications and utility industries. This situation positions service providers to sell excess capacity to other operators as well as explore new market opportunities. Trading and reselling can be a lucrative market opportunity. An intelligent optical network can enable bandwidth trading by connecting sellers and buyers at a bandwidth trading pooling point. In this application, an intelligent optical switch can become the workhorse for dynamically setting up short- and long-duration connections between bandwidth consumers and providers. An intelligent optical network would be required not only to manage optical channel connections, but will also work in conjunction with the multiservice switch/router to ensure QoS to meet service-level agreements (SLAs) between a seller and buyer.
These advances-higher bandwidth transport and intelligence-are emerging in the optical core of today's networks. Fortunately for service pro viders, multiservice switching and routing are also evolving to allow networks that can handle multiple types of traffic reliably, securely, and with a range of QoS levels.
With the growth of new services such as real-time multicast and VPNs, in conjunction with a demand for DiffServ and faster, cheaper access links, multiservice networks must accommodate QoS, network security, reliability and availability requirements, and evolving traffic-management protocols such as the MPLS standard. By providing support for IP, ATM, and MPLS with a multiservice switch/router, service providers can simplify networks and deliver QoS for new revenue-generating services over the higher-bandwidth intelligent optical core.
To uphold SLAs for these new revenue-generating services, providers must be able to guarantee that a customer's traffic is traversing the network as promised in the agreement. QoS-enabled transport for deploying DiffServ is needed to provide a level of predictability and control beyond best-effort IP services. For the service provider, QoS will mean new revenue potential, enabling tiered offerings to customers. This must be achieved with forwarding and real-time QoS controls implemented in hardware and transmission links allowing traffic to travel over multiservice networks at OC-192 (10-Gbit/sec) rates and beyond. Security and policy-based controls will also be essential to SLAs for preferred and differentiated services. Security measures such as VPNs continue to evolve as more enterprises move their traffic from private to public networks. Without QoS and security, SLAs would not be effective and DiffServ will not become a viable means of revenue generation.
Reliability and availability are also key elements, not only to support the massive growth in traffic, but also as networks shift from private to public networks and from traditional voice networks to multiservice networks. Capabilities like virtual routing will also aid the scalability of networks as multiple tunnels can be multiplexed into one larger flow of traffic. Furthermore, traffic can be classified as coarse or fine flows at ingress points, adding to the scalability of the network. All of these examples will become much more effective as the network layers converge and become more interoperable.
Traffic management is a key benefit of multiservice networks, which today leverage ATM technology. To support the same IP services over the multiservice network using an IP-based backbone, similar traffic-management features are required to offer efficient utilization of network resources via resource reservation and load balancing. In MPLS, this utilization is achieved by mapping the tunnels, or label-switched paths (LSPs), to current physical paths. Through MPLS traffic engineering, these physical paths use explicit routing and aggregate traffic in appropriately sized flows. The mapping takes into account the traffic conditions and resource availability in the network. Resources and priority are assigned to an LSP to support QoS, significantly diminishing the possibility of a telecommunications traffic jam. MPlS within the optical network allows it to create new physical paths to relieve congestion, wherever needed.
With MPLS, multiple hierarchies of tunnels can be established to terminate in various parts of the network, easing processing requirements of core network elements. MPLS enables fast, ATM-like label switching to speed up IP packet forwarding with small changes to existing routing protocols such as open shortest path first and intermediate system to intermediate system. With standardization of the interworking of MPLS control planes between data and optical switches, MPLS tunnels will be established between core multiservice switch/ routers over an optical network, simplifying service providers' physical networks and futureproofing the networks in place today. Leveraging MPLS, core multiservice switch/ router equipment can provide an alternative method for providing QoS for IP services over the optical core.
Combining support for new technologies such as MPLS, along with ATM switching and IP routing technology, the core multiservice switch/ router allows service providers to utilize "service agnostic" networks and flexibly deploy the technology best suited to the service requirements of end users. Delivery may be done using ATM technology-such as in the case of ATM constant-bit-rate service requirements-or IP/MPLS technology for IP delivery beyond best-effort IP.
The traffic-engineering and QoS capabilities for IP services offered by multiservice switch/routers, along with the advances in the intelligent optical domain, are making a significant impact on network topology plans and the overall evolution of combined data and optical networks. These multiservice optical networks must evolve to rise to the challenges presented by service providers.
Today, networks are built with multiple devices, each providing support for a specific network function. Specialized devices are provided for IP routing, Layer 2 switching, and so on. In addition, electrical signals are converted to light, and back again, as the signals travel between numerous data and optical elements. Traffic typically traverses SONET crossconnects, SONET add/drop multiplexers (ADMs), and DWDM equipment. ADMs are often required to link DWDM equipment to the slower interfaces of conventional switches and routers. Furthermore, all of the layers-routing, switching, SONET, and optical-are managed separately, which impacts overall operational cost (see Figure 1).
What's required for the converged multiservice optical network of tomorrow is a consolidation of this multilayer model into a simpler model consisting of multiservice switching/routing, intelligent optical networking, and end-to-end service provisioning (see Figure 2).
Greater integration of multiservice functionality with optical capabilities is in its early stages, for example, in the metropolitan area where multiservice switching/routing products promise to support IP/MPLS, ATM, time-division multiplexing (TDM), SONET, and DWDM. A new breed of multiservice switch/router is emerging with OC-192 capabilities that offers direct connectivity with DWDM equipment to ensure optical interoperability, removing the management complexities associated with the need for multiple layers of equipment. Direct DWDM connection simplifies operations, reduces power requirements, and can reduce equipment costs significantly with the elimination of optical translator units. These devices are likely to integrate TDM and SONET/SDH functions, as well.
Standards development is also taking place that will allow the multiservice and optical layers to work hand-in-hand. For example, dynamic optical signaling standards are evolving, allowing user devices such as multiservice switch/routers to dynamically request bandwidth from the optical network. That allows transport facilities to be created and canceled in an on-demand fashion. Combined with the new intelligent services in the optical layer, such as optical-layer interworking, these developments will clearly help lead to a more seamless network from the edge through the core.
In addition to the growing interoperability between multiservice and optical layers, network management is a critical concern of service providers. The explosion of bandwidth requirements combined with pressure to quickly fill new connection requests is placing great stress on the transport's network management for service provisioning, resulting in new requirements for network management that must be addressed. As services need to be turned on faster, network-management simplification is required to expedite service delivery.
The result is a need for self-managing network elements and the cost-efficient bundling of network management, network planning, and service management. Because of network evolution and burgeoning optical and multiservice switching capabilities, integration of optical- and data-layer management systems is a paramount concern. These new capabilities result in a fundamentally evolved telecommunications-management network (TMN) architecture, where the management-system "pile" is much shorter as the result of a new functionality distribution, although the functionality of TMN is still valid and respected.
Competition and new service providers alike drive capabilities such as self-planning and switched broadband connections to the individual network devices. To be competitive, the TMN architecture evolves in a number of ways. First, connection management is handled by the network elements in a distributed manner, moving the network-management functionality down to the network element. This approach opens the opportunity for providing fast switch-like connection setup and fast restoration. It also decreases the number of interfaces involved in the introduction of new equipment, bundling the connection management of the network element within itself. It eases handoffs and time lags between different systems such as server and client systems.
Therefore, deployment cycles become shorter, especially with new-product introductions. The interface to the network element becomes simpler and only the end-to-end service needs to be provisioned and can be done from a central workstation. This approach requires the device to have topology knowledge and forces the network's auto-discovery capability down to the network element. With a self-learning network, the risk of human error is decreased as are the staffing needs.
Service management is pushed down, too, bundled with a management system associated directly with each network element. This approach allows the network provider to turn on revenue-generating services much faster. Add to that the capability of providing switched transport connection, and new revenues from bandwidth trading and reselling are possible. Since the interface to the network elements is simpler, the service-management function becomes more generic and can be focused on the service aspects. This approach supports centralized, generic service centers, providing an integrated view across multiple technologies, if needed.
Furthermore, the element-management function is significantly simplified. It becomes mainly a storage function for performance-monitoring history, alarm management, and inventory/software management. This function can become more generic, too, independent of the nature of the data.
Finally, the management complex associated directly with the network element provides full network and network-element displays for the network provider as well as a subset of the network view for end-customers subscribing to a transport VPN. In addition, it provides a point for interface concentration, allowing the management system to have a single-and simpler-interface point, rather than having to interface with each device. This management complex can be duplicated for reliability purposes.
It's an exciting time for service providers to take advantage of rapid changes throughout the network. Not only are optical switches increasing the capacity of core networks, a multiservice switch/router infrastructure is allowing service providers to offer new revenue-generating services.
Furthermore, new capabilities in network management and intelligent devices are easing network-management strains and allowing operators to turn on services faster. The evolution of the intelligent optical layer, multiservice switching/routing, and network management for expedited service provisioning is revolutionizing the telecommunications industry.
Patricia Streilein is senior product manager, internetworking systems, and Joe John is a distinguished member of the technical staff, advanced optical networking, at Lucent Technologies (Murray Hill, NJ).