Many community fiber broadband projects seek to create a platform for competition among several service providers (SPs). "Open access" separates infrastructure – which tends to be a natural monopoly – from competitive services. SPs are expected to compete by differentiating themselves from their rivals on price, terms, customer service, brand, customization, or innovation, all to the benefit of consumers. The network provider (NP) operating the community's network wholesales commodity connectivity – "dumb pipes" – to the SPs so they can reach residential and small and mid market business customers.
Open access in fiber to the premises (FTTP) networks can be created via various forms of physical or virtual unbundling. In physical unbundling, the NP wholesales only the physical media – dark fiber or wavelength – to the SPs. In virtual unbundling, the NP provides access transmission and aggregation.
Many people incorrectly assume that only physical unbundling can fully achieve open access. We'll explain the implementation of several approaches and their effect on policy objectives, business models, capex, and operations.
A short history
The history of unbundled DSL has lessons for FTTP. In the late 1990s, regulators in several countries attempted to foster competition by unbundling. The loop unbundling model gives competitive providers direct access to copper metallic pairs from central office or pedestal to customers, along with collocation of transmission equipment and interoffice facilities. It initially was targeted at business phone services offered by competitive local exchange carriers (CLECs). Later, as Internet access over DSL emerged, ISPs and CLECs began to offer Internet service over unbundled loops. In principle, this model allows many forms of competitive differentiation.
In practice, there were numerous technical, operational, and legal complications with loop unbundling. Collocation logistics led to numerous disputes and regulations. DSL introduced unanticipated challenges; line conditioning, binder group integrity, loop qualification processes, and vectoring all were complicated by crossed lines of accountability. Customers served by remote electronics (digital loop carriers and remote DSLAMs) were a particular problem; "sub-loop unbundling" added further complexity and contention.
In cases when loop unbundling was impractical or not desired by SPs, "platform unbundling" (also called bitstream unbundling) applied. For DSL, that meant customers were served by a DSLAM and Broadband Remote Access Server (B-RAS) that the NP operated. Point-to-Point Protocol over Ethernet (PPPoE) over ATM-encapsulated Ethernet, PPP over ATM (PPPoA), or Layer 2 Transport Protocol (L2TP) provided password authenticated access. The authenticated user ID mapped the PPPoE session to the customer's selected SP.
Thus the NP provided a Layer 2 service between the customer and SP. The Internet service was branded by the SP, which provided a DSL modem, customer service, billing, email, and other services. Provisioning and trouble reports flowed between the SP and NP. The Broadband Forum (formerly ADSL Forum) Technical Report TR-025 specification describes that in detail.
Starting in 2001, the U.S. regulators exempted cable – and later DSL and FTTx – from unbundling. Arguably, that encouraged cable operators and telcos to bring broadband services to 92% of the U.S. population. It also caused the decline of DSL CLECs.
Implementing open access for FTTx
The key to open access architectures is locating the crossconnection between customers and SPs. Several approaches are shown in Figures 1-4. (The various colors denote resources dedicated to one SP, while the gray denotes NP resources shared by all SPs.)
|Figure 1. Approach 1: Fiber loop unbundling.|
Approach 1. The most obvious open access architecture is similar to loop unbundling (Figure 1). The SP supplies an optical-network terminator (ONT) and home gateway to each customer. The NP furnishes homerun dark fiber from the customer premises to the central office (CO). Service providers' collocate their optical line termination (OLT) equipment at the CO. Crossconnections between customer-dedicated fibers and SPs' OLT ports are made at a fiber distribution frame in the CO. The fiber frame may be manual or robotic or supplemented with an optical crossconnect (OXC). Only homerun, point-to-point architectures, typically Gigabit Ethernet, are supported in this approach.
|Figure 2. Approach 2: Fiber sub-loop unbundling.|
Approach 2. An architecture similar to sub-loop unbundling permits the use of cost-effective PON technology (Figure 2). Crossconnections between customer-dedicated distribution fibers and SP-dedicated splitters and feeder fibers are made in a remote node. Automated fiber crossconnects are preferable to manual ones to avoid frequent truck rolls and accountability issues. They require electrical power.
|Figure 3. Approach 3: Wavelength unbundling.|
Approach 3. Loop unbundling can be done at the wavelength level in PONs (Figure 3). Emerging time wavelength division multiplexing (TDWM) PON technology can support up to four SPs over conventional splitter-based PONs. Each SP's OLTs is assigned one of four pairs of upstream and downstream wavelengths. ONTs tune themselves to the wavelength pair associated with the SP. OLTs must be coordinated.
|Figure 4. Approach 4: Virtual unbundling|
Approach 4. Virtual unbundling (Figure 4), like platform unbundling, is done with Layer 2 aggregation, using IEEE 802.1 Virtual LANs with Provider Bridging. The NP is solely responsible for ONTs and OLTs, in addition to the fiber infrastructure. Aggregated traffic is transported between each SP and the OLT, typically over 1-Gbps or 10-Gbps Ethernet links. Collocation is not necessary. Details are described in Broadband Forum Technical Report TR-156.
Proprietary DWDM PONs that assign each customer to a wavelength pair at installation will not be discussed here.
There are several factors to consider, depending on the approach.
Capex. Capital cost and financing of new infrastructure is an overriding planning concern and places a premium on cost-effective architectures. Sharing of network elements is key to design efficiency – many resources, each used exclusively by one customer, are many times more expensive than many customers using one resource. The principal motivation for PON is resource sharing.
Approach 1 requires point-to-point homerun fiber from the customer to the fiber frame in the CO. No fiber is shared, so large-count cable and large fiber frames are needed. Depending on distance and take rate, point-to-point can be 10% to 120% more expensive than PON. Robotic fiber frames or OXCs are an additional cost. In approach 2, each SP has one or more PONs for each serving area, thus reducing feeder fiber count. Competition causes low take rates for each SP, resulting in less efficient sharing and requiring additional spare fibers. Robotic remote fiber frames or OXCs add cost, including the need to install electrical power. TWDM PON, used in approach 3 is not a fully mature technology. Tunable ONT components are presently expensive. Equipment cost is reported to be between 5 and 8 times that of TDM PON. Approach 4 has little or no incremental capital cost. Crossconnection is performed in the Layer 2 switching equipment required for all FTTx architectures.
Installation, provisioning, and churn. Customer adds, moves, and changes are a large operating expense and occur frequently with open access. Under all of the approaches, these activities involve populating various databases and configuration files. The approaches differ primarily in complexity of information and work flow at the crossconnection between customers and SPs. If ONTs are provided by the SP, a change also requires a truck roll. Since proper installation procedures are critical for reliability and eye safety, self-install is inadvisable.
Approaches 1 and 2 require crossconnection by rearrangement in fiber frames (or through an optical crossconnect). Manual frame operations are labor-intensive and error-prone and risk dirtying or damaging connectors. Jumpers become tangled and records inaccurate. Approach 2, in addition to these liabilities, requires a truck roll. Robotic fiber frames or optical crossconnects remove manual operations from the work flow – at a price. Approaches 3 and 4 changes involve only ONT and OLT port configuration and can be performed without human intervention.
Accountability. Accountability across organizational boundaries was an Achilles' heel of copper loop unbundling and remains a problem for FTTx. The problem can be summarized as complex analog and mechanical interactions across ragged boundaries. Physical security, personnel access, and workflow raise large numbers of operational, logistical, and legal concerns. Harmful interactions can occur between coexisting transmission systems. Troubleshooting often becomes finger pointing when multiple organizations are responsible for different physical layer elements. Thresholds for physical infrastructure performance and timely work-order execution are often contentious and lead to allegations of anticompetitive behavior. Delineation of obligations in tariffs, contracts, and regulations is complex.
These issues most affect approaches 1 and 2 since they involve mechanical and optical interconnection. Approach 3 is also affected because it requires coordination between competitors' OLTs. Approach 4 establishes a well characterized digital boundary at Layer 2 between providers, making the access physical layer the sole responsibility of the NP. It also doesn't require OLT collocation, and Layer 2 performance is relatively easy to monitor and enforce.
Scalability. The policy goals of open access are best achieved when many SPs compete for the customer's business. All of these approaches limit the number of competing ISPs that the infrastructure can support.
Approach 1 is limited by congestion at fiber frames, approach 2 is limited by rack space in remote cabinets, and approach 3 currently supports four wavelength pairs, allowing for no more than four ISPs. The VLAN tags used in approach 4 are limited to 4,092 service instances – one or more per SP; support for hundreds of ISPs is feasible. Approach 4 allows customers to have multiple SPs, for example, they might have different Internet services for household and home office use or might buy Internet and phone service from one provider and TV from another. This possibility is precluded by the first three approaches.
Competitive differentiation. Historically, SPs used innovation in transmission systems to differentiate themselves by performance and customization.
Approaches 1 and 2 enable SPs to select transmission systems based on their own criteria, while approach 3 enables them to select among the combinations of upstream and downstream line rates in the TWDM PON standards. In approach 4, the NP provides uniform Layer 2 services to the SPs, who can't use the physical and data-link layers for differentiation. However, transmission systems innovation is now channeled through industry standards rather than proprietary products. SPs can more easily differentiate at the IP layer in support services (like DNS and CDNs) and bundled services like TV, phone, email, and storage.
The goal of open access is to provide a level playing field for competing SPs. All four approaches meet that objective. Approach 4 does so in the most cost-effective and operationally efficient fashion, albeit at a small cost in opportunity for differentiation. Therefore, virtual unbundling is usually the best way to achieve open access.
Virtual unbundling resonates with the trend toward virtualization in IT and networking. By abstracting the physical infrastructure from commodity connectivity and transport, it provides a stable, robust, and flexible platform for competitive value-added services. Virtual unbundling is also an application for network function virtualization and software-defined networking.
Regulators in the U.K., New Zealand, and Australia have adopted virtual unbundling and structural separation of wholesale access. It appears to have a strongly pro-competitive effect: Wellington, NZ, has 16 competing SPs using the Chorus fiber network, and BT Openreach provides wholesale service to more than 300 competing SPs.
Open access with Layer 2 "dumb pipes" is the smart way to foster services competition.
Dan Grossman is a broadband technologist, FTTx expert, and principal of NetAccess Futures, a consultancy that helps broadband providers make cost-effective system technology planning and procurement decisions. He can be reached at email@example.com.