Imagine if, instead of building a network of paved interstate highways to support the advent of automobiles, we had tried to adapt the existing railway system for cars. Sounds crazy, right? However, operators of today’s global telecommunications networks face similar choices, with the same inherent problems.
The optical communications networks in use today have done an outstanding job of transporting the conventional voice and data traffic for which they were designed. However, more than half of the traffic traversing core long-haul optical networks now is packet-based. Analysts also expect global IP traffic to increase at a compound annual growth rate of more than 53% over the next 5 years (see Figure 1).
As worldwide usage of IP voice, video, and other applications continues to grow, service providers expect that IP will ultimately become the unifying standard for all traffic traversing their networks. To prepare for this IP-driven future, they have made major investments in technologies that enable the transport of packet traffic over core long-haul optical infrastructures that were initially designed for conventional voice and data services. However, like the railway scenario, the strategy suffers from the fundamental cost and inefficiency issues that one would expect when adapting an infrastructure to do a job for which it was not originally intended.
Transporting IP traffic across conventional core long-haul optical networks requires an extensive series of conversions. Carriers must convert IP packets into an intermediate electrical form, which is then transformed into optical wavelengths, transported across the core long-haul optical backbone, and finally converted back to standard IP at the other end of the network. (Within the networking industry, this series of conversions is referred to as optical-electrical-optical, or OEO.) In practice, this means adding multiple layers of equipment and operational processes at several points along the core network — and taking on the substantial costs and complexity that go with them.
Most service providers recognize that this situation is far from ideal. Fortunately, carriers have options for improving on their existing network architectures: They can transform their network cores by integrating IP and optical elements more closely within the network and eliminate much of the intermediary technologies and operational processes as well as their associated costs. This option, which can be referred to as IP over DWDM (IPoDWDM), provides more efficiently integrated IP and optical networking and supports unprecedented data rates for voice, video, and data services for decades to come.
To understand why carriers have reached the crossroads previously described, it helps to understand how today’s core long-haul optical networks came to be. Digital communications networks were originally designed and optimized for the transport of basic low-bandwidth voice phone calls. The entire network — including the optical backbone — was built to support the resiliency, reliability, and manageability of voice services.
Core long-haul networks were built as optical rings based on SONET or SDH transmission systems. (The systems are similar; SONET is used primarily in North American and Japanese telecommunications networks, while European and other Asian carriers generally use SDH.)
These transmission systems offered a number of advantages. In the first place, SONET/SDH provides self-healing capabilities that allow the network to very quickly (within 50 msec) reroute voice connections if one path along the optical ring fails. SONET/SDH networks also provide a standard industry interface for accommodating handoffs among various local, long-distance, and international carriers. Finally (and most critically for many service providers), SONET/SDH provides a variety of operations, administration, management, and provisioning (OAM&P) features that carriers use extensively to provision, protect, and monitor their optical networks.
When IP applications began to emerge in the early 1990s, service providers naturally instituted a model of converting packet traffic for transport over the existing SONET/SDH infrastructure, allowing them to extend SONET/SDH resiliency and manageability to new IP services. However, with the advent of the World Wide Web, carriers soon found their core long-haul networks inundated with IP traffic. To meet this need, DWDM technologies entered the scene, enabling the transmission of multiple SONET/SDH optical wavelengths over a single fiber — and freeing up huge amounts of bandwidth in the network core.
While this model does require the transmission of IP traffic over SONET/SDH transport, it has been effective for several years. However, in today’s world of IP-based television (IPTV), streaming music, podcasts, voice over IP (VoIP), and high-definition video-on-demand (VoD), operators are struggling to cost-effectively accommodate ever greater volumes of IP traffic in the network core with conventional SONET/SDH architectures. As new applications spawn the use of 10- and 40-Gbit/sec interfaces, and other interfaces that don’t easily map to SONET/SDH (such as 10-Gbit/sec Fibre Channel and 100-Gbit/sec Ethernet), the limitations of a core architecture designed for 51-Mbit/sec voice circuits become increasingly apparent.
To support the massive growth in IP traffic in the network core, carriers are exploring a new type of optical network backbone — one inherently optimized for packet-based services. This “next-generation” network core should meet three key requirements:
• First, it should provide greater capacity and scalability, without requiring frequent upgrades to the deployed optical infrastructure as bandwidth demands grow.
• Next, it must provide native support for Ethernet — increasingly the protocol of choice for IP traffic — without requiring complex and costly conversion processes.
• Finally, it should offer faster speed to service by allowing carriers to quickly, easily, and remotely scale core bandwidth with changes in traffic flows.
Carriers have two primary choices for “IP-optimizing” their network cores. They can continue building out electrical switching fabrics that convert IP traffic for optical transport using conventional SONET/SDH techniques. Or, they can adopt the IPoDWDM model to accommodate new growth and implement a fully optical switching network that efficiently converges the IP and optical elements in the core of the network (see Figure 2).
Today, many telecommunications carriers still maintain large electrical switching fabrics to convert core IP traffic for transport via SONET/SDH (and, ultimately, via DWDM). However, the approach presents a number of challenges.
First, an electrical switching model is expensive to maintain and grow, because it requires extensive OEO conversion equipment. As carriers add more devices and complexity to the network, maintaining optimal network uptime to meet the stringent service-level agreements they’ve negotiated with their customers becomes increasingly difficult. Since the electrical conversion equipment in the network core is often designed for a single type or speed of traffic, it is also more difficult and expensive for carriers to adapt this equipment as core bandwidth demands change with newer technology (for example, the emergence of 100-Gbit/sec Ethernet).
On top of these concerns, electrical switching technologies also have practical ramifications that reduce the network’s efficiency and increase its capital and operational costs over time. For example, a DWDM circuit entering a point-of-presence (PoP) in a carrier’s core network may contain both IP traffic that must be processed by routers at that location, as well as “pass-through” traffic that is simply on its way to the next PoP. In a modern core network, pass-through traffic can make up as much as 80% of the overall traffic traversing a PoP.
In the conventional electrical switching model, pass-through traffic must be demultiplexed and converted through a series of transponders at each PoP to support continuous SONET/SDH monitoring, error correction, and protection/restoration functions. However, these functions are actually redundant: Modern IP and Multiprotocol Label Switching (MPLS) protocols can provide the same monitoring and protection features (for example, through MPLS Fast Re-Route), eliminating the need to convert pass-through traffic to SONET/SDH entirely.
For carriers, this unnecessary processing of pass-through traffic is more than a minor inconvenience. The electrical switching equipment required to perform this conversion (at every single PoP) is very expensive, quickly eats up limited real estate, and requires significant additional power and cooling, increasing the service provider’s overall operating costs.
As network services continue to converge toward a single IP standard, many carriers (especially newer entrants to the telecommunications market, such as cable operators) are considering the IPoDWDM approach. Instead of employing distinct network infrastructures to provide optical transport (referred to as Layer 1, or L1) and IP routing and switching services (referred to as Layer 2/Layer 3, or L2/L3), the IPoDWDM approach converges components within the network elements, control-plane mechanisms, and management models of these network layers to create a more integrated, streamlined network core (see Figure 3).
Adopting a model that integrates optical and IP elements offers carriers a number of advantages. Obviously, by eliminating the need to maintain an extensive electrical switching system (as well as that system’s associated housing, cooling, and operating costs), the IPoDWDM model can reduce overall network costs and complexity. The model also streamlines the network core, since an IPoDWDM infrastructure can allow pass-through traffic to simply traverse PoPs without extensive, unnecessary OEO conversions.
IPoDWDM also provides greater optical provisioning flexibility, because packet traffic can remain in the optical domain through the entire core network until L3 IP processing is required. Since pure optical transmission is inherently less sensitive to lower-layer protocol changes than electrical transmission techniques, this transmission approach is essentially protocol agnostic. IPoDWDM technologies can provide native support for the full range of optical and IP protocols, including 10-Gbit/sec Ethernet, 40-Gbit/sec packet-over-SONET, and even 100-Gbit/sec Ethernet, which will support future applications.
The IPoDWDM approach also enhances a carrier’s ability to scale core bandwidth with changing traffic demands. With IPoDWDM, each PoP in the network core need not support multiple electrical transponders for each traffic type and speed. As a result, increasing core bandwidth becomes a simpler matter of upgrading the routing equipment interfaces at each end of a link, instead of upgrading the electrical switching equipment at each intermediate point across that link or network.
IPoDWDM also helps carriers manage service networks more efficiently. In conventional core networks, operators must transport IP traffic over SONET/SDH networks to support end-to-end OAM&P functionality. The IPoDWDM model employs a technology called WDMPHY, which uses the G.709 standard to add a “digital wrapper” of OAM&P information to IP traffic transported over the DWDM network. This additional layer enables service providers to extend the full range of OAM&P capabilities over Ethernet (and enjoy the cost advantages of Ethernet) without requiring SONET/SDH at all. In addition, when carriers integrate IP and optical services, they gain the option of using a single, simplified management platform to manage both the IP and optical elements in the network. This management integration can even extend to people and processes, since a converged management platform can support the integration of traditionally separate operational units managing optical and IP domains.
Looking ahead, this approach also will support a future network that can dynamically add core bandwidth (and additional wavelengths) as needed. While today’s SONET/SDH networks provide some dynamic bandwidth functionality, it is limited and must be performed in fixed 51-Mbit/sec increments. Tomorrow’s IPoDWDM model will allow carriers to scale in increments of 10-Gbit Ethernet or 40-Gbit Ethernet without having to completely reengineer their networks.
While IPoDWDM offers the potential to transform the way service providers grow and manage their core long-haul networks, this approach is more than just an interesting vision for the future. New IP and optical technologies enable carriers to begin employing this model today in their core networks.
For example, carriers are increasingly deploying advanced optical provisioning tools such as reconfigurable optical add/drop multiplexing (ROADM) components. ROADM technologies give carriers the ability to add or drop services and wavelengths as needed without modifying the physical fiber infrastructure. Service providers are also increasingly using core IP routing systems with 10- and 40-Gbit/sec tunable DWDM lasers. These interfaces allow operators to change the type of service offered over an existing fiber network without a wholesale equipment upgrade. In some cases, these routing systems allow the transmission of 40-Gbit/sec traffic over existing 10-Gbit/sec optical wavelengths. New optical mesh switching devices are also helping to pave the way toward all-optical, any-to-any port switching of optical signals. All of these technologies are laying a foundation for an optical core architecture that accommodates IP traffic in its native state, with all of the associated cost and efficiency advantages that implies.
As the price of advanced optical components continues to fall, IPoDWDM technology is being deployed in a growing number of telecommunications networks. Concurrently, as new applications such as IP-based high-definition VoD and IPTV-on-demand enter the market, telecommunications networks will continue to see dramatic growth in core IP traffic and evolve toward a converged core infrastructure based on IP. While telecommunications carriers are still determining the best approach for managing this transition, the IPoDWDM model should become increasingly attractive.
Greg Smith is senior marketing manager of optical networking and Tony Sarathchandra is a product marketing manager for core routing at Cisco Systems (www.cisco.com).