A Reference Model for RBB
This section develops a generic RBB network—that is, a reference model that incorporates the key features of RBB networks. A reference model is useful in establishing vocabulary and pointing out parallel features among various technical approaches. In addition, a reference model can help identify points of congestion in end-to-end service delivery.
Our purpose is to identify the functions of three logical elements of an end-to-end RBB network: 1. Core Network, 2. the Access Network, and 3. the Home Network. The Core Network is covered in this chapter. The Access Network is covered in Chapter 3, "Cable TV Networks" ; Chapter 4, "xDSL Access Networks" ; Chapter 5, "FTTx Access Networks" ; and Chapter 6 "Wireless Access Networks." The Home Network is dealt with in Chapter 7, "Home Networks."
Useful reference models for RBB networks have been developed by the Digital Audio Visual Council ( Digital Audio Visual Council DAVIC and ATM Forum Residential Broadband Working Group ATMF RBB ). The DAVIC Reference model was originally written in DAVIC 1.0 Specification Part 02, System Reference Models and Scenarios, September 1995. The ATM Forum Reference model is presented in the organization's document (BTD-RBB-01.01, "Residential Broadband Baseline Document Draft," February 1997).
Figure 2-1 depicts the end-to-end schematic for the reference model. The Core Network, Access Network and Home Network domains are segmented in this way to emphasize boundaries of control.
Figure 2-1. A Reference Model for RBB Network
Core Network
The Core Network provides high-performance bit pumps to move bits between the content provider and the Access Network. It is the repository for network management and application servers, such as Web servers and video servers. Speed and manageability are critical in the Core Network. The principle functions of the Core Network are listed here:
Switching, routing, and transmission
Handling concentration of traffic to the content provider
Handling service registration for multiple content providers
Multiplexing and switching for multiple Access Networks
Providing navigation aids and directory services
Enforcing Quality of Service (QoS)
Performing load balancing
Caching servers on behalf of the content providers.
The Access Network may also perform some of these functions. For example, it is logical to place caching servers in the Access Network as well as the Core Network.
The protocol options to consider in the Core Network are IP or MPEG over Asynchronous Transfer Mode (ATM), packet mode over Sonet (POS), and IP over Dense Wavelength Division Multiplexing (DWDM).
Figure 2-2 illustrates protocol stacking for the ATM approach, which has the (possible) current advantage of being conventional wisdom.
Figure 2-2. ATM Core Network Protocol Stack
In this approach to Core Network design, IP and MPEG packets from the application are segmented into ATM cells at the edge of the Core. ATM virtual circuits (VC) guide the packets through the Core to the intended destinations. Key issues of ATM point mainly to signaling overhead and the complexity of mapping IP addresses to ATM addresses. Call setup rates for ATM are still in the low hundreds per second, which puts ATM at a disadvantage with respect to pure IP networks. However, ATM does provide a partitioning capability so that multiple users share the same Core (unbeknownst to each other) with a high degree of QoS.
This figure also points out that ATM is layered on top of SONET/SDH networks, which in turn are layered on top of a fiber optic DWDM network.
POS advocates point out that the QoS and partitioning features of ATM can be replicated at the packet layer; therefore, there is no need for ATM at all.
However, some go further to say that SONET/SDH can be bypassed as well. SONET/SDH provides timing and error protection. But timing would be relatively less important for applications that can maintain strict timing control at the packet layer and that provide for a rerouting capability. Therefore, there is some discussion of an IP over DWDM architecture, which eliminates the cost of the ATM switches and the SONET/SDH add/drop multiplexers (ADMs) but potentially at some critical functionality loss.
These approaches are shown in Figure 2-3.
If sophisticated QoS schemes are implementable in packet mode (pure IP networks) cost-effectively, then the advocates of ATM and SONET/SDH will have some difficult questions to answer. Time will tell.
Access Network
The Access Network is the part of the carrier network that touches the customer's premises. The Access Network is also referred to as the local drop, local line, or last mile.
The principal functions of an Access Network are listed here:
Transmitting, switching, routing, and multiplexing traffic from the consumer to the Core Network
Classifying traffic from consumers by QoS; that is, the Access Network differentiates best-effort traffic from traffic with guaranteed bandwidth
Enforcing QoS
Providing navigation aids and directory services
Caching servers on behalf of the content providers
Performing tunneling or packet encapsulation
Enforcing the MAC protocol
Enforcing packet filtering
Providing software updates to the Home Network
Authenticating users
Providing measurements used for invoicing
These functions are performed cooperatively with the Core Network and the Home Network, which means that there must be clear interfaces among these systems.
At Issue: Network IntelligenceAn important system design question is how network intelligence is to be optimally distributed between the Core Network and the Access Network. Intelligence here refers to processing of user signaling (for example, call setup requests). [Quayle] argues that the Access Network should be relatively dumb and the Core Network relatively intelligent. Also, large databases for invoicing and resource management should be maintained in the Core Network. Because there are relatively few Core Network nodes and a large number of Access Network nodes, total system costs are optimized by concentrating the expensive (intelligent) pieces in the Core Network. Others argue that because the Core Network must move bits as fast as possible, it must be simple, even dumb. Therefore, intelligence is pushed to the periphery, even to the customer premises. The case for dumb networks (overprovisioned, underengineered networks) and smart peripheral equipment (and people) is argued most publicly by David Isenberg ). ATM switches, for example, are relatively simple, with most of the hard software functions performed outside the switching fabric. However, because some Access Networks require a Media Access Control protocol, there is a requirement to interpret user signaling in the Access Network. Hence, for these networks, there is some requirement for some intelligence in the access. Whether or not both Core and Access Networks will be dumb remains to be seen. Finally, on a commercial matter, it may be the case that the Access Network and the Core Network will be different companies. Each will assert that intelligence should be put into its network, which boosts added value. |
Many types of Access Networks are discussed widely in the popular and trade press. Table 2-1 shows a list of the Access Networks in operation or in contemplation, with associated standards for telephony, video, and data. The acronyms will be discussed in due course; we only attempt to indicate here which Access Networks could support which broad class of service.
Though multiple types of Access Networks exist, there are three architectural elements, which they share. These are the access node, the Optical Network Unit, and the Network Termination (NT in Figure 2-1).
The major aforementioned Access Networks are covered in subsequent chapters. HALOs, blimps, and powerline are viewed as too experimental as of this writing. However, powerline transmission deserves some mention because of its potential.
Powerline is the use of electric utility wires to transmit data. Because computers and networking equipment need electricity, it makes sense to consider using the electrical outlet for a data connection as well. The issues with using powerlines have revolved around electrical emissions, noise on the lines and cost. Also, it is not possible to pass data through a transformer. However, changes in transformer design are giving rise to the possibility of this new Access Network. Principal developers include Northern Telecom ), United Utilities of Great Britain, and Sydkraft of Sweden (in conjunction with the Swedish Internet Service provider, Tele2). Northern Telecom and United Utilities maintain a powerline Web site at www.nor.webdpl.com.
A typical U.S. power grid supports only four to eight homes per transformer, so powerline is viewed as expensive. But in Europe, each transformer supports 100 to 200 homes. A networking device at the transformer site is amortized over a larger number of homes, thereby making powerline more cost-effective. If noise and emission problems can be solved, this could represent an interesting alternative.
Access Node
The access node (AN) serves some or all of the following functions:
Modulating forward data onto the Access Network
Demodulating return path data
Enforcing the Media Access Control (MAC) protocol to arbitrate access from multiple users onto the Access Network
Multiplexing traffic from the Access Network to the Core Network
Separating or classifying traffic prior to multiplexing onto the Core Network, such as differentiating traffic that is subject to QoS guarantees from traffic that receives best-effort support
Enforcing signaling
Performing passive operations, such as splitting and filtering
Enforcement of the MAC protocol is one of the more complicated functions of the AN. For some Access Networks, when multiple consumers vie for bandwidth connections simultaneously, arbitration rules are required to determine who gets to use the network at that moment. Without some controls, the Access and Core Networks would be obliged to provide an indefinite amount of bandwidth and connections. The MAC protocol arbitrates access to the Access Network. It is enforced by the AN, which engages in a protocol exchange, or peers, with the NIU on the customer premises. Important MAC protocols are token-passing schemes, Slotted Aloha protocols, and Collision Sense Multiple Access protocols.
Examples of ANs are Cable Modem Terminal Server (CMTS) in the Multimedia Cable Network System (MCNS) architecture for cable data modems (discussed in Chapter 3), Digital Subscriber Line Access Multiplexer (DSLAM) for concentration (discussed in Chapter 4), and the ATM Digital Terminal (ADT), referenced in the ATM Forum RBB Baseline Document.
Optical Network Unit
The function of the Optical Network Unit (ONU) is to terminate fiber and convert the optical signal on fiber to electrical signals on wired networks. A major difference among Access Networks is the proximity of the ONU to the consumer. The closer the ONU is to the consumer, the higher the speed and usually the greater the service reliability. Carriers and consumers alike would like to have fiber as close as possible to the home—perhaps even inside the home—as service called Fiber to the Home (FTTH).
Network Termination
The Network Termination (NT) is a carrier-provided or customer-provided piece of equipment that is located on the side of the home. Some of its functions are listed here:
Coupling home wiring and carrier wiring
Grounding
Handling RF filtering
Handling splitting
Performing media conversion
Performing remodulation
Handling security and interdiction
Handling provisioning
The NT is the legal and commercial demarcation point between the Access Network and the consumer. This means that if something goes wrong on the network side of the NT, the carrier is obliged to fix it. The NT is situated on the customer premises, but many countries include it as a part of the Access Network.
In the United States, the NT is considered to be customer premises equipment because the customer purchases it. In Europe, it is considered carrier property. For purposes of this reference model, the NT is part of the carrier network because it usually is under carrier control, even though the customer purchases it. The fact that it can be purchased from a number of suppliers other than the Access carrier is not relevant architecturally.
NT functions differ widely among Access Networks, but the main functions are usually passive (coupling, splitting, grounding, RF filtering, and so on). In some cases, the NT can have an active function, such as signaling. For example, service activation might be enabled by signaling between the carrier and the NT. This becomes an important concept for Fiber to the Home (FTTH), which is discussed in Chapter 6.
Chapter 7 contains more discussion of various Home Network components.
Why Multiple Access Networks Exist
Proponents of specific Access Networks claim that one Access Network would fit all applications. While it is technically feasible that a single network could service all market drivers, there are reasons why multiple Access Networks exist:
Embedded base— Telephone companies and cable operators have put billions of dollars into the ground. This represents investment that cannot be easily replaced. The sensible thing is to leverage it.
Different applications— Broadcast television lends itself economically to shared media, such as air and cable. Retrofitting broadcast media to accommodate point-to-point applications involves software tricks. Similarly, retrofitting broadcast applications over point-to-point media also involves software tricks.
Different population densities— Some technologies are more economically viable in urban areas than rural areas, and vice versa.
Different geography— Vast expanses of oceans and rivers create a cost penalty for wired networks. On the other hand, obstacles such as hills, buildings, vegetation, and heavy rainfall reduce transmission quality for wireless networks.
Different business conditions— In developing countries, which are characterized by a lack of mature wired infrastructure, broadcast technologies can create a fast infrastructure. Regulations such as those imposing limits on market dominance of a single carrier can foster or inhibit some technical options.
Entrepreneurship— The availability of capital underwrites a lot of development of various technologies.
Because of these variations, subsequent chapters consider a variety of Access Networks to understand the strengths and weaknesses of each.
Home Networks
The Home Network consists of the following aspects:
The Network Interface Unit (NIU), essentially a modem
The Residential Gateway (RG), which adds network functionality
The Set Top Unit (STU), which performs application-specific functions, such as decoding digital TV
The Terminal Equipment (TE), a television, personal computer, or other device
Consumer premises distribution, either wired or wireless
In the home, the personal computer and consumer electronics industries will wage a spirited battle for the consumer dollar. PCs, TV set-top boxes (STB), and game stations will blur as each acquires some of the functions of the others.