Providing efficient, scalable access networks
Metro Solutions / SPECIAL REPORTS
Deploying equipment that links LANs to the metro network is the next expansion step for service providers.
KEVIN RYAN, Terawave Communications
The LAN, metropolitan core, and long-haul networks have been deployed. New technology and products have enabled service providers to build highly scalable and cost-effective solutions in the metro core and long-haul networks. Metro core and long-haul networks are now capable of carrying terabits of traffic at very low cost per megabit or service. LAN bandwidth in the enterprise easily scales with the emergence of 100Base-T and Gigabit Ethernet, available at very low costs. The missing piece to the puzzle is how to connect the LAN and the metro core while providing bandwidth scalability, service flexibility, and most important, profitability and revenue growth for the service provider.
The existing copper infrastructure in the access network was deployed long before the emergence of broadband network services. Today's copper plant was designed to handle plain old telephone services (POTS), not broadband services. Band-aids to the existing network have proven inadequate as a solution. DSL, in all its forms, still doesn't scale and has severe bandwidth, services, and distance limitations.
Moving SONET deeper into the network and deploying SONET nodes at the customer site is extremely expensive and inefficient for all but the top fraction of the customer base. These existing technological limitations of existing solutions have dictated a new access architecture paradigm to meet the explosive bandwidth growth in the network.
To solve the access bottleneck issue, service providers and system solution providers must first understand and work with the fundamental truths of the business access network, instead of shoehorning old technology into the space or putting Band-aids on existing infrastructure.
First, the access network is a point-to-multipoint architecture. In all business access networks, there is a hubbing point, whether it's a central office (CO), point of presence (PoP), digital loop carrier (DLC), or digital subscriber line access multiplexer (DSLAM). These hubbing points branch out to serve hundreds to thousands of customers. Second, there is a broad mix of services needed to meet customer requirements. It is ideal for both the customer and service provider to offer these services over a single infrastructure. Third, service providers must start a network small and scale it in a pay-as-you-grow manner, allowing quick payback of initial builds and sufficient headroom for growth as new subscribers come online. Finally, churn will always occur in the access network. The customer's service needs change over time as business requirements for networking services evolve (services churn). Also, businesses frequently change location. Today, a business location can be a dentist's office and tomorrow an Internet service provider (ISP) PoP (customer churn). The better the service provider can adapt to these two types of churn, the better the chance of keeping customer loyalty and scaling profitably to meet the needs of the customer.
There are three fundamental topological architectures for addressing the access network: rings, point-to-point, and point-to-multipoint (see Figure 1). To be efficient, rings require multiple nodes on the ring, which only works if the entire ring is built at once. Unfortunately, business customers in the access network do not buy services at the same time, and it is too costly to build nodes prior to customers signing up for services.
Given this reality, most rings in the access network are collapsed rings or single-node rings (see Figure 2). To add subscribers to the ring after the initial build, service providers must break the ring to add additional customers. That is very costly and complex, requiring service outages in unprotected networks. In essence, SONET rings in the access network become point-to-point solutions.
The main problem with point-to-point networks is their inefficiency when it comes to outside plant (OSP) issues. In these architectures, each building needs two fibers in unprotected configurations and four fibers in protected configurations. This approach may prove in for the small number of high average revenue per user per month (ARPUPM) customers, but does not scale down to meet the needs of the overwhelming majority of business subscribers in the access network.
For example, in the case of point-to-point Ethernet, it involves single-service, high-speed Internet traffic and does not address the other service needs of customers. In this approach, an entire overlay network is required to meet the full service needs of the customer. In the case of point-to-point SONET architectures or collapsed rings, costs are very high and there is a lot of stranded bandwidth in the network.
If these point-to-point alternatives are extrapolated down to the hundreds or thousands of subscribers supported by a CO or PoP, it's easy to see why a point-to-point solution does not scale to meet the needs of small and medium-sized businesses. For example, if a service provider wants to pick up 500 businesses in a wire center, it needs to terminate 2,000 feeder fibers in protected mode. As a quick comparison, to terminate the same number of businesses using a passive-optical-network (PON) approach, only 32 feeder fibers and CO/PoP terminations are required. These numbers quickly illustrate the inefficiencies of serving a point-to-multipoint access network with point-to-point solutions.
Point-to-multipoint solutions, such as PON, overcome the inefficiencies of the two architectures previously mentioned. A PON solution can serve multiple subscribers from one feeder fiber, enabling aggregation of services over a single fiber into the metro (see Figure 3). OSP requirements are reduced from 2,000 to 32 feeder fibers, reducing network terminations proportionally.
The second issue for service providers extending the metro core to the customer is diversity of services and churn in the access network. Customers in the wire center have varying requirements for services from POTS to T1 (1.554-Mbit/sec) connections and Ethernet and optical (see Figure 4).
The more services the provider can offer, the higher the ARPUPM. ATM PON, unlike other point-to-point or ring solutions, allows service providers to cost-effectively offer multiple services over a single infrastructure. Services are dynamically added to the customer site without the need to change infrastructure, customer-premises equipment (CPE), or CO/PoP termination equipment. This flexibility allows service providers to dynamically adjust services based on subscriber needs in a pay-as-you-grow manner.
The third issue concerning service providers is the efficient addition of more subscribers to the network (customer churn). In point-to-point networks, service providers must engineer the entire network-from the CO/PoP all the way to the subscriber-to add a single customer. SONET rings are even more complex to upgrade. Here, the service provider must break the ring or create a collapsed ring to support the new services.
This lack of scalability to pick up additional subscribers in the service area makes it very costly and time-consuming to add users to the network. PON greatly increases the ease of adding customers to the network. With PON, proximity customers to existing PONs can be added by simply building out the lateral from the split point, thus greatly reducing the cost of adding incremental customers to the network (see Figure 4).
This ease in scaling the network to support more users gives the service provider deploying PONs a huge advantage over other service providers deploying ring or point-to-point architectures. First, the PON-based service provider can turn up services quicker-only engineering the lateral, as opposed to the lateral, feeder, and CO/PoP. Secondly, their costs are much lower (lateral costs + CPE versus lateral + feeder costs + CPE + CO/PoP equipment). Adding incremental customers under the PON approach can easily be a fraction of the cost of other solutions, and payback for incremental customers can be five to 12 months, depending on the services and lateral construction costs.
Broadband service requirements are rapidly extending out from the metro to the suburbs and from the main business districts to the small-office/home-office (SOHO) and residential customers. The shortfalls of the existing network become even more problematic further away from the CO/PoP, which is seen by the sparse coverage of services such as ADSL. PON technology enables services providers to drive broadband services deeper into the network and all the way to residential subscribers (see Figure 5).
PON can serve as effective broadband delivery and aggregation of services from DSLAMs, DLCs, and wireless tours. By using PON as the delivery and back-haul medium, today's service providers can cost-effectively deliver highly scalable services deeper into the network, effectively extending the reach of the new broadband network deep into the service areas.
As the service requirements of business and residential users evolve, so must the network. The extent of the changes in bandwidth and services required dictate a departure from the existing strategy of putting Band-aids on existing infrastructure or forcing legacy metro solutions into the access network.
With advancements in LAN, metro, and core networks, the access network has become the last major bottleneck to providing scalable, cost-effective broadband services to business and home users. PON technology has been developed in coordination with leading carriers around the world to provide the answer to the last network bottleneck-the access network.
Kevin Ryan is director of product management marketing and business development at Terawave Communications (Hayward, CA). He can be reached via the company's Website, www.terawave.com.