One of the challenges to be faced when structuring an optical access network is the very broad spectrum of potential applications and the multiplicity of solutions being developed to meet its needs. Various network topologies can be successfully used to meet the needs of high-speed networking: hub and spoke, multidrop, ring, and mesh. The possibility of intermixing access network technologies within the network further complicates the situation. In the end, the performance and suitability of any combination of network configuration and technology depends on the bandwidth and scalability required as well as the nature of the current network, including any legacy systems, economic factors, and future expansion plans.

The characteristics of optical access networking are as follows:

• High data rates (up to several Gbps) are being transmitted over distances that are relatively short. The majority of access network links will be less than 35 km in length, with many much shorter than that. In the minority of cases, where new service providers are building networks with relatively few PoPs in a geographic area, links may need to be as long as 70 km. However, this is unusual.

• The most effective use of optical fiber in an access network is to carry information directly on individual wavelengths. While SONET technology has been used in some cases, its optimization for voice traffic multiplexing imposes penalties on its use for data transmission. Many modern optical access networks now use SONET–less connections between enterprises (CPE) and service provider premises (PoP or CO), minimizing cost and complexity.

• Physical layer-optical access (Layer 1) allows any protocol or service to be carried over previously unlit fiber. Physical layer connectivity is the most effective method of unblocking the bandwidth bottleneck between CPE and the CO or PoP using dark fiber solutions. Physical-layer access must still accommodate certain essential network requirements: manageability, flexibility, and affordability. Optimally, physical-layer support can be used for speeds that range from a few Mbps (e.g., T1) to several Gbps (e.g., optical carrier [OC]–48), including fibre channel and Gigabit Ethernet.

• Support will be offered for both wavelength division multiplexing (WDM) and non–WDM links—with network economics determining which makes most sense for any given application.

• All nodes in a network will have some form of processing capability—although it is likely that intelligence will be distributed around the network. Ideally, the more complex functions may be concentrated at a central node (e.g., a hub in a hub-and-spoke network) where system management can be focused.

• The most successful marriage of bandwidth and optical access network technology will make use of an embedded communications channel with a management interface. In other words, management information will actually be carried within a wavelength, alongside high-speed data without a bandwidth penalty. Also, given the dominance of signaling network management protocol (SNMP) as a de facto management tool, it can be expected to be the protocol of choice—although, over time, other management protocols may be required and will need to be supported.

One of the most important aspects of the optical access network is its potential to provide not simply high bandwidth, but also a high QoS with corresponding performance monitoring to maintain that quality. Until now, SONET has traditionally been used to provide quality monitoring. Although its capabilities in this area are solid, they are also costly and cumbersome for data traffic. Other techniques, such as digital wrapper, make use of management bits, symbols, frames, packets, or cells wrapped around user data but inevitably incur a similar processing and cost penalty. Wrapperless techniques are beginning to emerge that offer the opportunity to manage link quality and performance measurement effectively, without the overhead of wrapper solutions.

These new solutions can not only provide service level management (service level agreements [SLAs]) economically, but also offer the use of 3R (versus 2R) techniques—retiming, regeneration, reshaping—for signal integrity in all channels in each direction and plug-and-play, high-quality network solutions.