All-in-one or best-in-class: As the range of services continues to increase, segmentation by delivery systems offers carriers the strongest business case.
Optical networking in the metropolitan area has become a topic of increasing interest and debate, with proponents advocating a wide variety of solutions for mass deployment. These range from intelligent DWDM systems and next-generation SONET systems to passive optical networking and simple Gigabit Ethernet switches.
The breadth of proposals arises from the huge market potential in the metro and widely heterogeneous carrier requirements. For carriers to deliver a successful business case in support of mass metro deployment, vendor solutions must meet a plethora of specific technological demands. Services offered in the metro space range from narrowband DS-1 (1.544- Mbit/sec) private-line, ATM, or frame-relay services; to broadband Gigabit Ethernet connectivity carrying Internet Protocol (IP); to direct transparent wavelengths.
Figure 1. In a metro access network (a), at least 90% of the traffic is destined for somewhere outside the immediate subnet, such as a regional network, backbone network, or the Internet. In a metro core network (b), where the access networks are aggregated, traffic flow is more evenly distributed within the network.
The trend from startups as well as the "big players" in this field has been to integrate multiple elements and services into a single box. Typically, four to six network elements are condensed into a single unit, driven by the reduced space, power, and capital costs associated with an integrated platform. But as the range of services from a single platform increases, it becomes more and more difficult to maintain best-in-class status in each category. Carriers often face the difficult choice of using a single-vendor solution that meets most, but not all, of their needs and are compelled to include additional boxes to complete the solution. Furthermore, all-in-one boxes force a compromise between feature set and scale; for example, optimizing for T1 (1.554 Mbits/sec) instead of OC-192 (10 Gbits/sec), resulting in very different price points and feature sets.
With an increasing realization that the one-box solution cannot solve all networking needs, it is important to identify the natural segmentation points that lead to the best business-case results for metro carriers. Following are the best possible segmentation points and their resulting impact on deployment viability.
A number of popular segmentations have been proposed in metro environments to address the varying carrier requirements. A popular segmentation is the notion of metro core and metro access, in which the metro core consists of interconnection between carrier points of presence (PoPs) or central offices (COs), and metro access consists of connection from the carrier to the customer premises. Vendors may optimize their solutions to provide best-in-class products for either the metro core or metro access.
This two-tiered carrier segmentation is accurate for carriers such as incumbent local-exchange carriers, which have several COs throughout a metro regional area from which their customers can be accessed. However, many next-generation carriers do not have a two-tiered model, but instead have direct connections from the backbone PoP directly to multitenant-building complexes. In this scenario, the multitenant units (MTUs) look very much like carrier PoPs, and the distinction between customer premises and carrier PoPs is reduced. These subtleties reveal the difficulties associated with making clean segmentations of the metro market.
Figure 2. SONET and DWDM systems from different vendors are often managed and maintained independently as separate networks. In this scenario, SONET add/drop multiplexers are used for all electrical traffic multiplexing, grooming, and switching in narrowband applications that require an aggregate ring capacity of <1 wavelength. DWDM systems are used to add fiber capacity in broadband applications that require aggregate ring capacity of multiple wavelengths.
There is a distinction in traffic patterns between metro core and metro access. In a metro/regional network, the traffic tends to be more evenly distributed, whereas in metro access, the traffic is generally backhauled to a CO or "super-PoP" where it is connected to the national backbone. This approach works for both two-tier model carriers as well as single-tier carriers, whereby customers are directly accessed. Thus, the distinction between metro core and metro access can be made based on the underlying flow of bandwidth being distributed or backhauled. Figure 1a shows a typical example of traffic patterns in a metro access network, and Figure 1b illustrates traffic flow in a typical metro-core application. As general segments of a market are defined, the question becomes: What is the most effective method to deliver services for these segments?
Another segmentation that has achieved significant deployment is the separation of the DWDM layer from the SONET or other switching/multiplexing layer. In this case, all electrical switching such as data or SONET multiplexing is done in a single box under one vendor's management domain, while a separate vendor delivers increased bandwidth through a DWDM system (see Figure 2). In this scenario, the metro-core market is dominated by DWDM-based equipment and the metro-access market is dominated by SONET technology. The advantage of such a division is that DWDM technology can be added where it is required and traditional rings can be used in fiber-rich situations. In some simple cases, a SONET vendor may even incorporate ITU wavelength transceivers on the line side of a SONET add/drop multiplexer (ADM) to reduce transponder costs.
The DWDM or optical layer not only reduces fiber count, but also incorporates readily available technologies such as optical bypass, eliminating unnecessary and costly hop-by-hop optical-to-electrical-to-optical (OEO) conversions for transit traffic. For high-bandwidth requirements, this savings is compelling in a fully meshed architecture in the metro core as well as in a star-over-ring architecture in the metro access.
For carriers delivering high-bandwidth services such as OC-3 (155 Mbits/sec) through OC-48 (2.5 Gbits/sec) to end-customers, DWDM thus becomes an important technology in the metro core as well as in the metro-access space.
The weakness of the SONET- and DWDM-based segmentation is that it treats the DWDM system as a fixed inflexible system functioning simply to increase fiber capacity. In fact, next-generation optical service provisioning platforms (OSPPs) are much more dynamic, using optical switches and wavelength tunable components to react to network demands and faults. For example, a single optical switch may protect hundreds of gigabits of traffic from fiber fault. The traditional SONET layer has no visibility into such switching schemes. In a client/server architecture where SONET ADMs are clients to the optical-networking layer, the network is left with separate control and management domains.
In addition, next-generation DWDM OSPPs use an IP or IP-like control plane such as Multiprotocol Label Switching to quickly provision and reconfigure network requirements to meet bandwidth-on-demand applications. These control schemes can be optimized to the mesh connectivity present in DWDM networks but are not optimized for the hop-by-hop SONET ring architectures. That leaves the SONET layer ill-suited to take advantage of the leading-edge features of OSPPs such as optical reconfigurability, protection, and switching. Although separation of a SONET layer from an optical layer is convenient for migrating legacy networks, it is not well suited for next-generation metro networks.
The cost and complexity associated with multiplexing in the electrical domain may suggest that pure optical systems have the advantage in providing metro services. An important consideration, however, is the capital cost associated with delivering such services.
Optical-networking equipment, such as next-generation DWDM systems, is cost-effective when the services delivered are greater in bandwidth.
For example, it is extremely cost-prohibitive to carry a single 45-Mbit/sec DS-3 line using traditional wavelength multiplexing, since DWDM systems have a fixed cost associated with a signal that is independent of bandwidth (the fiber and filters). It costs the same to pull off a wavelength of OC-192 from a DWDM ring as it does an OC-3 wavelength, which naturally makes OC-192 service attractive and OC-3 less attractive.
Figure 3. Segmentation between large-scale bandwidth systems based on optical DWDM technology or optical service provisioning platforms (OSPPs) and smaller-scale bandwidth systems based on multiservice SONET provisioning platforms (MSPPs) achieves the greatest scale with the widest service breadth. In locations where narrowband DS-0 and DS-3 services are required, an edge-multiplexing platform can be deployed; where direct broadband OC-48 and greater services are required, a scalable optical-based platform-OSPP-can be deployed; and where both types of services are required, the trunk of MSPP can be fed as a client to the OSPP.
This fixed-cost argument also applies to pure dark-fiber multiplexing, where it is very cost-prohibitive to directly map a DS-3 line to a fiber when multiple DS-3 lines are to be carried. It is important to note that pure dark-fiber implementations lack the protection, monitoring, and manageability available using SONET or DWDM technology, so they cannot be viewed in an apples-to-apples comparison with other multiplexing schemes. SONET and DWDM technologies not only offer the protection, monitoring, and manageability necessary for many metro services; but they also provide the ability to multiplex multiple signals onto fibers.
In the case of SONET technology, this signal/wavelength can then be shared among multiple sites on a ring, maximizing the return on a single wavelength. But on the other end of the scale, it becomes very inefficient to multiplex signals using SONET technology when the bandwidth requirements are very high. Only four OC-12s can be multiplexed onto an OC-48 wavelength, giving only a 4-to-1 multiplexing gain over direct dark fiber.
In contrast, tens or even hundreds of OC-12s can be multiplexed onto a single fiber using DWDM technology. Based on this argument, SONET multiplexing technologies are cost-effective when delivering lower bandwidth services from OC-3 and below but are ineffective for higher-bandwidth chunks such as nxOC-12/48/192 services or Gigabit Ethernet and Fibre Channel services.
To improve low-end efficiency, next-generation DWDM-based systems may also incorporate a thin layer of SONET multiplexing to deliver lower-end services such as OC-3s and OC-12s. This multiplexing may become cost-effective, since these so-called "thin-SONET" solutions can be implemented on a single chip. Such DWDM systems can then be very cost-effective at delivering services from OC-3 up to nxOC-192. In contrast, next-generation SONET systems such as multiservice provisioning plat forms (MSPPs) can be very effective at aggregating lower rate signals from DS-0 to DS-3, and possibly OC-3, into higher-band width signals such as OC-12 and OC-48.
These edge-pro visioning plat forms, from simple SONET ADMs to full multiservice platforms with support for ATM, IP, and frame-relay protocols, can provide rich service definitions for narrowband services. Since the bandwidth scale remains relatively small on such platforms, greater service flexibility and more interfaces can be introduced. With narrowband interfaces covered under a single platform, the optical system can focus on higher-bandwidth services such as SONET OC-12/48/192, Gigabit Ethernet, and wavelength services with a narrower set of service definitions (see Figure 3).
Next-generation metropolitan-network deployments offer huge opportunities and challenges for incumbent and competitive carriers. Time-critical selection of vendor solutions must be completed to compete for new customers and opportunities.
However, successful deployment means justifying the right business case for different service types. All-in-one solutions are very attractive on paper but often lag in feature set and time-to-market versus more focused solutions.
The correct segmentation of service delivery becomes critical to a scalable and effective solution. Traditional segmentations of either metro core and metro access or the SONET layer from the optical layer have many pitfalls.
A proposed alternative is the segmentation of large-scale bandwidth systems based on optical DWDM technology, or OSPPs, and smaller-scale bandwidth systems basedon multiservice SONET pro visioning platforms.
This segmentation achieves the greatest scale with the widest service breadth. In locations where high-bandwidth services are required, a scalable OSPP can be deployed. In locations where narrowband DS-0-to-DS-3 services are required, an edge-multiplexing platform can be deployed. In cases where both types of services are required, the trunk of the MSPP can be fed as a client to the OSPP.
To achieve maximum service flexibility for carriers, the management and provisioning schemes of the two platforms should interoperate seamlessly. As a result, MSPP players and OSPP vendors, working together, can deliver a complete solution for mass metro deployment.
Near Margalit, Ph.D., is the founder and chief technical officer for Zaffire Inc. (San Jose, CA). Margalit is also the founder of the Zaffire product architecture. For more information, see www.zaffire.com.