More capacity in the optical access network-but how?
With the technical evolution of the optical core progressing, carriers must weigh the pros and cons of available optical access solutions.
Jerry Power and Kent Novak
Networks are dynamic entities that evolve as customers' needs evolve and as new technology becomes available. The technical evolution of the optical core is relatively straightforward: Add more optical capacity. Operators are moving to higher transmission rates and adding DWDM systems with greater channel counts to support more capacity per fiber. The dramatic increase in wavelengths in the optical core has created the need for photonic crossconnects that will displace traditional time-division multiplexing (TDM) systems and relegate them to the roles associated with sub-lambda management.
While the direction of the transport systems in the core is clearly all-optical, the technical evolution of the access network is not as obvious. Leased-line services (predominantly 1.5 Mbits/sec and above) continue to grow at a significant rate. ATM access, primarily driven by asymmetric digital subscriber line (ADSL), is increasing dramatically, and the projected growth rate for Internet Protocol (IP) is phenomenal.
Furthermore, while out-shadowed by data services, traditional voice services for fixed lines and wireless subscribers are also experiencing rapid growth. While ADSL is increasing the capacity of the residential and small-business subscribers through the use of existing copper infrastructure, it does not address the service needs for large businesses and value-added service providers. More often, these customers are requiring solutions to handle high-bandwidth requirements, perhaps even a dedicated optical channel, in a cost-effective manner.
All of these services require more capacity in the access network, but due to differing quality-of-service (QoS) requirements, they cannot (or should not) be handled in an identical manner. Ideally, each service should be carried in a transport format that maintains the expected QoS, while still capitalizing on its inherent efficiencies, such as statistical multiplexing.
A variety of solutions for the optical access network are being proposed in the marketplace. These solutions include offering pure wavelength services, having IP carried over optical fiber, proprietary multiplexing over fiber, and SONET/SDH multiplexing (sometimes used in conjunction with DWDM). Each of these solutions has advantages and weaknesses when compared to the alternatives.
The simplest optical access architecture runs a dedicated fiber to each carrier customer. The fiber is offered on a dim or dark basis and implies that the customer is given the fiber as a physical entity with no restrictions on its use. While this approach is easy to visualize and administer, it uses fiber capacity extremely quickly and precludes the carrier from pricing services based upon use.
The idea with direct optical wavelength access architecture is to provide a dedicated wavelength to each customer rather than an entire fiber. Though providing a dedicated wavelength to a customer, typically via DWDM, does not address the issue of pricing services, it does significantly increase the number of wavelength services that can be offered over a single fiber.
Today, DWDM systems can be deployed as point-to-point links, linear add/drop runs, or wavelength add/drop devices in ring architectures (see Figures 1 and 2).
The dedicated wavelength architecture has some advantages over other access architectures. For instance, if a customer needs to upgrade service requirements from OC-3 (155 Mbits/sec) to OC-12 (622 Mbits/sec) or OC-48 (2.5 Gbits/sec), only the customer-premises-equipment (CPE) devices need to be upgraded. Additionally, carriers can offer customers non-SONET interfaces such as the Escon interface used between mainframes and computer storage farms. Finally, dedicated wavelengths have good security characteristics-wavelengths are difficult to "tap" into.
On the negative side, dedicated wavelengths are obviously best suited for large-capacity services and cannot cost-effectively transport lower-speed leased lines via dedicated optical channels. Additionally, this architecture relies on optical (analog) performance monitoring which, to date, has been inferior to digital performance monitoring and does not provide a solid basis for QoS or service-level agreements.
Ideally, all communications services could be supported with a single protocol. At one time, Integrated Services Digital Network (ISDN) was lauded as "the convergent technology," and currently IP with Multiprotocol Label Switching (MPLS) appears to be the focus of such discussions.
Once this technology becomes readily available, it will allow further consolidation of the data network. As data services dominate voice services, the intent is to economically absorb voice networks into this same service architecture. Thus, some equipment vendors propose a network where all telecommunication services are converted to IP at (or very close to) the customer premises and transported throughout the network in this native format.
The clear benefits of an IP-over-optics solution are the network efficiencies achieved via statistical multiplexing in data applications. However, statistical-multiplexing techniques become bandwidth-intensive once traffic levels exceed engineered targets, which could result in increased delays and error rates. Statistical studies from various universities indicate that packet networks should never be engineered for more than 60%-75% utilization.
The fallout of such factors is that, while multiple services could be equalized and carried over a common network with data traffic, such a move would significantly increase the transport capacity required to support the current voice network.
Leased-line services, which continue to grow considerably, add another dimension to the IP-over-optics approach. Customers purchasing leased lines from carriers not only expect guaranteed bandwidth between two points, but also believe they are purchasing a "clear pipe" that will not be manipulated. For years, that capability has been readily provided by circuit-based infrastructure. In more recent times, the performance characteristics of leased lines have been improved via enhancements of SONET/ SDH. Now, advocates of IP proclaim that all services, including leased lines, can be efficiently handled in an IP network. But since the economic benefits, or penalties, of handling traditional leased-line services via an IP service infrastructure re main inconclusive, it is unclear whether these "cut-through" services will ever be fully consumed into the IP/ MPLS network or remain ancillary.
There are proprietary alternatives to a standards-based multiplex layer (SONET/SDH) or service multiplexing (IP). Proprietary systems are not burdened with the requirements of a standard system and, as such, short cuts can be taken to target an efficient solution for specific problems. Proprietary multiplexing solutions may be implemented at the optical level such as a 4:1 optical mux/demux capability or at the service layer via specific cell or packet multiplexing techniques.
Since proprietary multiplexing is often targeted toward a specific application, it is typically the most cost- effective solution for that specific application. An example in the optical domain is a proprietary 4:1 optical mux/ demux unit used to transport wavelength services. While this technology is very cost-effective at multiplexing and transporting 4xOC-3 wavelength services, it does not accommodate higher-bandwidth services, nor is it the most cost-efficient method for transporting bandwidth in general.
Additionally, unlike a standards-based solution, a proprietary multiplex system must be "mirrored" by an equally proprietary demultiplexer. For example, a SONET-partitioned wavelength can be directly fed into a crossconnect, while a proprietary partitioning scheme must be demultiplexed into its constituent components before it can be passed to the crossconnect.
Finally, most operations and test systems are designed with the expectation of a standards-based approach. Operations are made more complex and costly with proprietary systems that require specialized test equipment.
A wavelength is currently capable of supporting a 10-Gbit/sec digital signal, and the technology exists to soon handle 40 Gbits/sec. However, few end-customers need an entire wavelength, making wavelength services inherently inefficient. A standards-based (SONET/SDH) multiplexer provides a mechanism to fan-out a single wavelength to multiple customers (see Figure 3).
SONET and SDH evolved as industry standards for optical signal framing and performance monitoring across an optical link that would allow multivendor interoperability.
In this regard, SONET and SDH must be considered a huge success. SONET and SDH are the predominant technologies deployed around the world to support transport applications, with tens of billions of dollars worth of equipment being deployed every year. However, this overwhelmingly predominant transport technology does have its limitations.
While SONET/SDH equipment provides carriers with end-to-end management, granular bandwidth for various services, and a standard for interconnectivity, it is inherently a circuit-based technology. Traditionally, this characteristic translates as "inefficient" for data services. More specifically, a 52-Mbit/sec SONET "pipe" delivered to a customer is always connected to that customer. It does not matter whether the customer is transmitting 52 Mbits/sec, 10 Mbits/sec, or zero data: the dedicated connection still exists. Furthermore, SONET/ SDH bandwidth (52 Mbits/sec, 155 Mbits/sec, etc.) is not well aligned with transmission rates commonly associated with data bandwidth (10 Mbits/ sec, 100 Mbits/sec, etc.).
Finally, while the SONET/SDH system can easily increase the bandwidth allocated to a particular service up to the total aggregate capacity of the system (i.e., OC-48 or OC-192), any expansion beyond the aggregate capacity requires replacing all systems in the optical path. This problem has been mitigated via DWDM systems working in conjunction with multiple SONET/SDH systems to increase the total capacity of the existing fiber infrastructure (see Figure 4).
A review of the technologies and solutions outlined here could leave a carrier in a real quandary as to which solution should be used as the basis for the access network architecture. SONET/SDH is a proven technology that can be deployed in a managed multivendor environment, but it does not easily accommodate the rapidly growing data demands. Although IP is efficient for data traffic, does it provide a viable solution for the existing leased lines and voice services as well as the performance standards currently demanded? A proprietary multiplexing solution can be optimized to a carrier's specific applications, but what are the hidden operational costs and what limitations are introduced as network interconnect is attempted? Finally, DWDM costs are rapidly decreasing, yet dedicated wavelength services are only efficient for the grandest bandwidth requirements.
In attempting to resolve the question of "which solution is best" it is helpful to prioritize the issues to be addressed and allow for alternatives that are not mutually exclusive with respect to technology. More specifically with respect to the priorities, the tremendous increase in data traffic has significantly increased the total capacity requirements of the access network. However, a carrier must be capable of delivering all services in a reliable, managed way. Furthermore, increased competition demands a standard multiplexing scheme that can easily operate in a multivendor, multicarrier environment. Finally, increasing the capacity of the embedded fiber infrastructure will allow carriers to compete more profitably.
Fortunately, some telecommunications-equipment vendors have realized this combination and are providing optical multiservice nodes (OMSNs) that transmit data-oriented protocols such as IP, ATM, or Ethernet over a standard SONET/SDH infrastructure. The result combines the benefits of statistical multiplexing for data services and the proven operational capabilities of SONET/SDH.
An OMSN generally consolidates similar traffic types onto a virtual tributary or virtual path as a SONET/SDH signal (see Figure 5). Described in very basic terms, a wavelength can be thought of as a pipe, wherein SONET/SDH breaks that pipe into a series of smaller partitioned pipes. Each smaller pipe can carry a different service protocol. Service protocols divide the bandwidth within the pipe into individually billable service segments. For example, an STS-1 (52 Mbits/ sec) can be used to carry many packets-each of which could be sold as an independently billable unit of service.
The OMSN is equipped with "native" Ethernet, ATM, or IP service interfaces as well as traditional DS-1 (1.544 Mbits/sec), DS-3 (44.736 Mbits/ sec), STS-1, OC-N transmission interfaces. Thus, interconnection costs between the CPE and access transport equipment are minimized. Performance monitoring and path restoration are performed by SONET/SDH, thus ensuring the high QoS level expected for all service types. Finally, OMSN systems can either have integrated DWDM or work in conjunction with standalone DWDM systems to provide multiple wavelengths over a single fiber.
Although IP over SONET/SDH may be viewed as the best solution for now, as technology evolves and the network requirements change, so will the preferred solution. But when will this happen and what is the resulting network evolution?
In time, the QoS capabilities of IP/MPLS will increase. Additionally, the performance monitoring and protection mechanisms of the optical network will be significantly enhanced via the introduction of "digital wrappers" in DWDM systems and optical crossconnects. The resulting effect will be the capability to transport a larger number of services effectively and reliably in IP/MPLS directly on the optical layer without the use of SONET/SDH (see Figure 6).
When that is accomplished, IP over optics will likely be the most versatile, economic access transport solution. However, SONET/SDH systems will still be used in many cases to provide a multiservice transmission platform or cost-effectively transport leased-line services. In essence, it is anticipated that the access network will continue to be a multiservice environment with SONET/SDH, continuing to play a prevalent role, along with IP and DWDM.
Jerry Power is senior director for transport network infrastructure at Alcatel USA (Plano, TX). Kent Novak is Alcatel USA's vice president of marketing for transmission systems. Novak can be contacted at: firstname.lastname@example.org.