Plans for Ubiquitous Broadband Access to the National Information
Infrastructure in the Ameritech Region
Statement of the Problem
The national information infrastructure (NII) can be visualized as consisting of four components: (1) access transport from the premises of the end user to an access node; (2) access nodes on an intracity, intercity, and international network of networks; (3) the internodal network of networks, including the internetwork gateways; and (4) information repositories. The access nodes, network, and information repositoriesthe last three componentsare common facilities; their costs are predominantly proportional to usage and can be shared among many users. In contrast, the access transportthe first componentis a dedicated facility connecting the premises of a single user; its cost is predominatly fixed, independent of usage, and borne by that single user.
As a result, the access transportpopularly referred to as the "last mile," although, in fact, it is typically a few miles in lengthpresents the greatest challenge to the provision of a ubiquitous, truly "national," wideband NII. Because the access nodes and information repositories can be provided modularly and grow with usage, they are already being put in place, by small entrepreneurs as well as by large service providers. Since the intracity, intercity, and international traffic can be multiplexed into the traffic stream on the existing telecommunications network, that portion of the ''information superhighway" already exists. The major investment hurdle, then, is the deployment of ubiquitous broadband access to individual user premises.
At first glance, it might appear that cable television systems constitute such broadband access, but closer analysis reveals that this is not so. Cable systems are predominantly one-way downstream, some with minor upstream capability, usually relying on polling to avoid interference among upstream signals. More constraining, they are broadcast systems, with 50 or so video channels being simultaneously viewed by several thousand users. The same cost hurdle described above exists to upgrade these systems to provide dedicated broadband access to individual users.
Currently available technologies for access transport span a wide range of capabilities and costs. Since the costs cannot be shared among users, at the low end of the range they are shared among usages, particularly with voice telephony. Modems provide dialup capability over standard telephone lines at speeds typically up to 9,600 bits per second, with less common capability at 14,400 and 28,800 bits per second. Integrated services digital network (ISDN) is rapidly becoming widely available, providing dialup capability at 56, 64, and 128 kilobits per second. Various local carriers, including the local exchange carriers, are offering frame relay service at 56 kilobits per second and 1.5 megabits per second, and switched multimegabit digital service at 1.5 and 45 megabits per second, with plans for migration to asynchronous transfer mode (ATM) at OC-1 (45 megabits per second) and higher SONET rates. However, all of these require a dedicated leased digital line between the user's premises and the carrier's switch of a capacity equal to the peak bit rate used. As a result, these higher-speed services are cost effective only for large user locations that generate sufficient traffic to fill them. The challenge, again, is to provide cost-effective, high-speed access to individual homes and small business locations.
Deployment Plans for the Ameritech Region
Ameritech has committed to providing broadband access in its region, using a hybrid optical fiber and coaxial cable architecture that can support a wide range of applications, including video services similar to current cable television, expanded digital multicast services, interactive multimedia services, and high-speed data services. By providing a platform for a wide range of applications, a cost-effective solution is achieved (economic data are provided in a later section).
Construction is expected to begin in 1995 in the Chicago, Cleveland, Columbus, Detroit, Indianapolis, and Milwaukee metropolitan areas, ramping up to an installation rate of 1 million lines per year by the end of 1995 and continuing at that rate to deploy 6 million lines throughout the region by the end of the decade. In December 1994, the Federal Communications Commission granted approval for Ameritech to construct the first 1.256 million lines, distributed among the metropolitan areas as follows:
The system architecture employs a hybrid transport network of optical fiber and coaxial cable. Signals are delivered over Ameritech's ATM network to video serving offices, each serving 100,000 to 150,000 customer locations. The signals are then distributed on optical fiber to individual nodes, each serving a total of 500 customer locations, not all of whom may actually subscribe to the service. From each node, the signals are distributed on four parallel coaxial cable systems, each serving 125 customer locations. With this architecture, the coaxial cable network is less than 2,000 feet in length and contains, at most, three amplifiers to any customer location.
The signal on both the optical fiber and the coaxial cable is a broadband analog video signal. The initial deployment will have a bandwidth of 750 megahertz, with capability for upgrade to 1 gigahertz when the reliability of such electronics becomes proven, yielding 110 channels of standard 6 megahertz video bandwidth. The allocation of these 110 channels to various applications is flexible and will be adjusted to satisfy user needs. Based on current estimates, approximately 70 of the channels will carry analog video signals for applications similar to current cable television, including basic and premium channels and pay-per-view. The remaining, approximately 40, of the channels will be digitized using 256 quadrature amplitude modulation, yielding a usable bit rate of over 36 megabits per second on each channel. Approximately 30 of these digitized channels will be used for multicast services, with multiple users viewing each transmitted program. Approximately 10 of the digitized channels will be used for switched interactive services, for which each user requires a dedicated digital circuit for the duration of the session.
On the digitized channels, the video signals will be compressed using the MPEG-2 compression standard. Depending on the particular application, each such signal will require a fraction of the 36-megabit-per-second or greater capacity. The signals will be multiplexed at the video serving offices and demultiplexed by the customer premises equipment, using the MPEG-2 transport layer protocol.
In addition to the downstream capacity, the system will have an upstream capability provided by up to 20 channels, each of 1.5-megabit-per-second capacity. Depending on local conditions of noise and interference, it is expected that at least 15 of these will be usable on each coaxial cable system serving 125 customer locations.
The system is intended to be a platform for a wide range of applications. Accordingly, the customer premises equipment may be a set-top box for use with a television set or an adjunct to a personal computer.
As described above, each branch of the distribution network will consist of an optical fiber to a node serving 500 customer locations, with four coaxial cable buses from the node, each serving 125 customer locations. Each branch will have a downstream capacity of approximately 360 megabits per second for switched interactive services.
The system is designed for economical growth in capacity as the market expands and traffic increases. Initially, each fiber branch will share a linear laser in the video serving office with two other fibers, through a passive optical splitter, so that the 360-megabits-per-second downstream capacity will serve 1,500 customer locations, not all of whom may subscribe to the service. When the traffic grows to exceed this capacity, the fibers can be driven by independent lasers so that the 360 megabits per second downstream capacity will serve 500 customer locations. When the traffic requires it, up to four fibers can feed each node, one for each coaxial cable bus, so that the 360 megabit per second downstream capacity can serve 125 customer locations.
The downstream capacity will be assigned to the users on a per-session basis, depending on the particular application. There is some uncertainty about the bit-rate required for each application. Human factors experiments and customer trials appear to indicate that live programs other than athletic events, compressed in real time and displayed on full-size television screens, can be delivered at 3 megabits per second. Material that is prerecorded and stored in compressed digital form can be delivered at a lower bit-rate, since the compression is not performed in real time. During the compression process, the results can be observed, and the parameters of the compression algorithm can be optimized to the particular material. Similarly, video material that accompanies text and occupies only a portion of a computer screen requires less resolution and can be delivered at a lower bit-rate. Not all interactive applications will involve video, and those that do will generally involve material that is stored in compressed digital form and displayed in a window occupying a portion of the screen, so that an average of 3 megabits per second per session is probably a conservative estimate. At that average bit-rate, the 360-megabits-per-second downstream capacity could serve 120 simultaneous sessions.
Further, although the initial system will assign a fixed bit-rate for the entire duration of the session, equal to the bit-rate required for the most demanding segment, the capability will exist for statistically multiplexing the signals and using a bit-rate based on instantaneous requirement. This capability will be employed if necessary. In that event, an average bit-rate per session of 3 megabits per second would be quite conservative.
It is estimated that, during the peak usage period of the day, 15 percent of the subscribers will be using the service at the same time. That would consume the entire 120-session capacity when 800 of the 1,500 customer locations became subscribers, equal to 53 percent market penetration. At that point, the next step in capacity growth, sharing the capacity among 500 customer locations, would be implemented. If the peak usage per subscriber turned out to be higher than the 15 percent estimated, then the capacity growth would be implemented at a lower number of users. This would not affect the economic viability of the system, since it is the total amount of usage that will generate both the need and the revenue to support the increased capacity.
At the limit of the current architecture, the 120-session capacity will be shared by the 125 customer locations on each coaxial cable bus. That would support 96 percent simultaneous usage at 100 percent market penetration. If that turned out to be insufficient, because of multiple simultaneous users per customer location, or higher bit-rate applications, statistical multiplexing could be employed.
In addition to the downstream capability, each coaxial cable will support at least 15 usable upstream channels at 1.5 megabits per second each, and these will be multiplexed onto the fiber from the node to the video serving office. Therefore, unlike the downstream capacity, this upstream capacity will be shared by 125 customer locations from the start. These upstream channels must utilize a time division multiple access protocol, which does not allow for 100 percent "fill"; nevertheless, traffic studies indicate that a single 1.5-megabit-per-second upstream channel could easily support all of the digital video requirements, including video on demand, of the 125 customer locations, leaving the remaining 14 or more for the more interactive multimedia and data applications.
There are two parameters of the interactive services that determine the required upstream capacity: (1) the frequency (and size) of the upstream messages generated by the user and (2) the required latency, or speed of response to the message. Analyses of the types of applications that are anticipated indicate that the second
parameter is controlling. If enough upstream capacity is provided to assure that the transport delay contributes no more than 100 milliseconds, it does not matter how frequently the user generates upstream commands.
Of course, if the users were to generate large files for transmission in the upstream direction, rather than the inquiry and command messages typical of interactive services, this analysis would not apply. But, by definition, such users would be generating sufficient traffic to use one of Ameritech's other, more symmetric, data offerings cost effectively, and would not be using the hybrid fiber/coaxial cable system.
At a 100-millisecond latency, each upstream channel could support approximately 20 simultaneous sessions, for a total of at least 280 simultaneous sessions on the 14 or more upstream channels. This is greater than the 125 simultaneous sessions supportable by the downstream capacity that will only be generated by 96 percent simultaneous usage at 100 percent market penetration.
At this time, the FCC has approved construction of the first 1.256 million lines, and the economics for that initial phase, which is extracted from Ameritech's application for construction approval, is presented below.
Table 1 presents the economic data for the first 10 years. It is important to note that these data are strictly for the transport network; they do not include any costs or revenues for the provision of "content."
Construction of the first 1.256 million lines is planned for completion early in the third year. Both market penetration and usage per customer are expected to grow throughout the period, as shown by the revenue forecast. These revenues are for the transport of all types of content, including broadcast television, video on demand, interactive multimedia services, and high-speed data access. The total costs each year consist of three components: (1) the costs of constructing additional lines, which ends in the third year; (2) the costs of adding equipment for additional customers and additional usage per customer; and (3) the costs of providing customer service.
The cash flow each year, revenues minus costs, is discounted to Year 1 equivalent values and cumulated, and is presented in Table 1 as cumulative discounted cash flow (CDCF). As shown, the CDCF turns positive in the eighth year. In actuality, as additional lines are constructed in each of the six major metropolitan areas, building on the initial base, economies of scale are expected to make the economics even more favorable.
The greatest challenge to the realization of a ubiquitous wideband national information infrastructure, an NII that is truly "national," is the provision of economical broadband access transport to individual residences and small businessespopularly referred to as the "last mile." Access transport to large business locations can be based on usage, because such locations, as well as the nodes and internodal network and the information repositories, are all shared facilities and are used (and paid for) by multiple users. By contrast, access transport
to individual customer locations must be sized for the peak requirements of the most demanding application, even though that capacity is unused most of the time, and its cost must be borne by that individual customer.
Although these facilities cannot be shared by multiple customers, they can be shared among multiple services to the individual customer. Ameritech has committed to providing broadband access in its region, utilizing a hybrid optical fiber and coaxial cable architecture, supporting a wide range of applications. These include video services similar to current cable television, expanded digital multicast services, and interactive multimedia services, as well as high speed data services. By providing a platform for a wide range of services, a cost-effective solution is achieved.
Ameritech plans to begin construction in 1995, ramping up to a rate of 1 million lines per year by year end and continuing at that rate to deploy 6 million lines throughout the region by the end of the decade. The FCC has granted approval for construction of the first 1.256 million lines, in the Chicago, Cleveland, Columbus, Detroit, Indianapolis, and Milwaukee metropolitan areas.