`Last-Mile` Bandwidth & Conference BOF
[ Top of Page ]
[ Introduction ]
[ `Last-Mile` Bandwidth & Conference BOF ]
[ Survey Results ]
[ Author Information ]
Although bandwidth and computing capacity are issues throughout the Internet, it is
perhaps most acute in the connections people have into their Internet Service Provider
(ISP). It is the one parameter, other than content, which figures most heavily in their
perception of the Internet. It is also about the only aspect, other than the capacity of their
own computer, that they can actually do something about. In the following sections we
take a look at current technologies, survey technical issues and review policy issues
associated with the new, high-bandwidth data communications services.
Computer professionals all know that response time is a critical aspect of user-computer
access. Although it has been often reported, it has been most recently described by Jakob
Nielsen (Nielsen, Jakob, "User Interface Directions for the Web", Communications of the
ACM, January 1999/Vol. 42, No. 1, pp. 65-72 and http://www.useit.com/) in relation to
the WWW. He describes necessary response times as: less than 0.1 second for the
appearance of instantaneous feedback, less than 1.0 second for the user's flow of
thought to stay uninterrupted and 10 seconds as the limit for keeping the user's
attention focussed on the dialogue.
Why is this so important? It is clear that high speed communications provides a more
positive graphics experience on the Internet. But just the ability to download graphics
faster is not the primary benefit. One way to enable more people to use the Internet
effectively is to present it as a graphical experience, both in terms of content and user
interface which requires response times as described above.
Two examples come to mind. All of the new models for health care assume that individuals
will take greater responsibility for their own interactions with health care professionals. The
key to this is access to medical records. Imagine if these records were all on line (with
suitable security and privacy protections) and patients could access data in their record
by graphically exploring this data by looking at a chart showing the timeline of significant
events. See the LifeLines project of Ben Shneiderman and colleagues at the University of
The second example is help for the problem we have all faced of being "lost in cyberspace".
How many times have you worked your way through a web site (or series of web sites) only
to discover that you had no idea of where you were in the structure or how to get back to
a previous point? Although many sites try to provide some textual information on location,
what's missing is the computer equivalent of the familiar "You Are Here" maps found in
many physical environments such as museums, zoos, etc.
Why don't these capabilities exist? They do, but not on the Internet. And they don't exist
on the Internet because the bandwidth needed to provide the greatly increased amounts
of data simply is not available to most users. In fact many who access the Internet over
dial up lines do so with the graphics initially turned off!
For a summary of the potential, see the first SIGGRAPH public policy white paper
(SIGGRAPH White Paper, "Computer Graphics, Visualization, Imaging and the GII:
Technical Challenges and Public Policy Issues",
Because our knowledge is primarily of the situation in the United States, this description is
of necessity oriented towards the U.S. However, some of the technologies described (e.g.
the G.lite version of ADSL) are being globally standardized by the International
Telecommunication Union (ITU) (http://www.itu.int),
and others (e.g. for cable TV networks) will
likely be adapted to conditions elsewhere.
Current and Near Term Future Technologies
Currently and in the near term future, few residences and small businesses will be directly
served by optical fiber, because the construction and equipment costs of this technology
can seldom be recovered within a reasonable time from such customers. Thus we don't
discuss "fiber-to-the-home". Starting in the 1980s, a growing number of fiber-based
competitive local telephone companies (telcos) were launched to serve businesses in
metropolitan areas. A few electric and/or natural gas utilities (including municipally owned
ones) also began to use their rights of way for new fiber networks. The 1996 Telecom Act
encouraged such trends, and this will gradually expand the number of available broadband
data services, often in combination with other technologies that we examine more closely
It should be noted that broadband services for the residential market are usually priced at
flat monthly rates between about US$40 and US$60, vs. typical current ISP charges for
dialup access of around US$20/mo. We feel that these prices are generally attractive to
consumers, given the improved performance.
Digital Subscriber Line
The telcos plus some ISPs and new local carriers using telco wires are starting to deploy
and offer Digital Subscriber Line (DSL) service in a number of metropolitan areas. DSL
employs a high frequency carrier signal on existing copper lines to provide high bandwidth
digital communications. DSL can coexist with normal analog phone service. DSL is a
continuously connected service. DSL connections will be supported by many of the usual
DSL comes in many variations, hence the acronym is frequently written as xDSL. The
following paragraphs describe the key parameters. Note that some forms of DSL are
asymmetric, meaning that the upstream bandwidth is lower than the downstream
bandwidth. Depending on the application, this may or may not be a problem. Except as
noted, potential data rates decrease on longer loops (with the supported limit being about
18,000 feet from the telco switching office). Line quality and configuration also affect
performance, and some providers impose speed limits for marketing reasons, e.g. to
enable tiered offerings priced for different customer groups.
1) HDSL: The earliest DSL service was developed by Bellcore in the 1980s. HDSL uses
two wire-pairs (all other xDSLs use one) to provide T1 or fractional T1 service at lower
cost than T1's original 1960s technology. Data rates are 1.5 Mbit/s in each direction. It is
now being superseded by newer technologies.
2) SDSL: Advanced DSP chips enable T1 service (like HDSL) to be transported on only
one wire-pair, so the limited capacity of telcos' copper cables can be used more efficiently
and costs lowered.
3) IDSL: IDSL provides "narrowband" ISDN's Basic Rate service (144 Kbit/s
bi-directionally for the full 2B + D channel configuration), transported via DSL technology
so as to lower the telco's costs.
4) ADSL: ADSL is an asymmetric service designed for MPEG-2 compressed TV
movies-on-demand, and is now used for Web surfing. It trades off upstream bandwidth
(capped between 640 Kbit/s and 1 Mbit/s, depending on the implementation) so that up to
8 Mbit/s can be sent downstream. The asymmetry also deters businesses from replacing
T1s with less-costly ADSL.
5) "G.lite" or "splitterless" ADSL: This is ITU-T's new standard for home Internet access.
Unlike all other DSLs (for which an installer must visit the customer's site), consumers
simply plug their "modem" into a home phone jack. To achieve this labor-cost and usability
benefit, data rates are capped at 1.5 Mbit/s downstream and 384 to 512 Kbit/s
6) VDSL: This still-evolving short-range service is ADSL's upward migration path. On a
1000-ft. loop (from a fiber terminal), up to the OC-1 rate of 52 Mbit/s can be sent
downstream, vs. 1.5 to 6.4 Mbit/s upstream. A symmetrical business version is to carry
the E3 rate of 34 Mbit/s bi-directionally. At VDSL's 4500-ft. loop length limit, speed
roughly equals ADSL's. Telcos plan "fiber to the neighborhood" in support of VDSL
wherever ADSL's use is heavy enough to justify the cost of running fiber. If multichannel
HDTV movies-on-demand become popular, that would also justify early VDSL rollouts.
Good places to find out more about xDSL are http://www.adsl.com,
Cable Data Services
Cable television companies are also rolling out their answer to the telephone companies'
xDSL services. A special interface (misleadingly called a cable 'modem') is used to
interface a computer to the cable for data services. These interfaces are typically
packaged as an external peripheral device which is housed in its own case like an external
dial-up modem. The interface to the computer is then via a 10 Mbit/s Ethernet.
Bandwidths are typically up to 10 Mbit/s downstream and are capped between 200 Kbit/s
and 2 Mbit/s upstream. Cable data service is a shared service for several subscribers
using the same channel and travelling through all these subscribers' interface cards.
Packets are encrypted as a privacy measure . It is also a continuously connected service.
Cable services, while regulated, are not common carrier services so Internet access is
typically provided by captive service providers, such as @Home or RoadRunner.
Good places to find out more about cable data services are http://www.cablelabs.com and
http://www.home.net or http://www.rr.com
Satellite television providers have some limited data services. For example, DirecPC
provides downstream data rates of 400Kbit/s. Because there is no practical way to
provide uplink capability to subscribers, upstream data services must utilize another
medium, such as telephone. Note that for some applications, such as Web surfing, the
upstream data rate requirements are quite low and could be satisfied by analog modem
Good places to find out more about satellite data services are http://www.direcpc.com and
http://www.loral.com -- especially the latter site's details on CyberStar and Loral Orion.
Cellular telephone service has had the ability to transmit data for a long time. But the
bandwidths are low and the costs are high, so the service has primarily been used by a few
individuals who wish to take advantage of Internet connections (mostly for email only)
Companies like Ricochet are now providing medium speed data connections by blanketing
an area with low power transmitters/receivers to provide an Ethernet-like connectivity
without wires. It is only available in a very few metropolitan areas and is used mainly by
travelling professionals who want Internet access.
High speed broadband data services are starting to be provided by companies such as
WinStar, Teligent and Advanced Radio Technology. The plan is to have fiber connected
local hubs which feature line-of-sight to nearby buildings. These services are provided on
frequency bands as high as 38GHz. Current economics result in targeting business
customers located in buildings of 10,000 square feet or greater. Consumer service is not
yet available, but multi-unit dwellings like apartment houses and condominium complexes
have costs of service comparable to businesses of similar size. Thus, homes of this type
are an attractive future market for service providers like those named above.
Other firms have been freed by the U.S.' 1996 telecom reform to offer broadband wireless
data services, and some have signaled interest in doing so. Among the potential players
1) Established wireline telcos, which have for years used wireless local loop (WLL) radio
equipment to serve very remote or inaccessible customers' premises. AT&T has been
actively developing a new generation of such technology.
2) Digital cellular and PCS carriers, who are standardizing a new "third generation" (3G)
family of services that carry data at up to the E1 rate of 2 Mbit/s. Note that fixed-site
cellular data service dates from the 1980s, when it came into use for telemetry in
scientific, industrial, and security surveillance applications.
3) "Wireless cable" television service providers, and other owners of radio frequencies
newly-auctioned by the FCC.
4) Established terrestrial TV broadcasters, who may use parts of their newly-granted
digital TV channels for data transmission to paying customers.
Good places to find out more about terrestrial wireless data services are
http://www.winstar.com and http://www.teligent.com.
Digital Subscriber Line
Digital subscriber line services have two major technical limitations: they are distance and
crosstalk limited. The telcos' plans for incremental extension of fiber into residential areas
(noted above under VDSL) will slowly address these constraints by shortening the copper
loops used for older forms of DSL. Just how big a problem these shortcomings will be really
waits to be assessed until the service is more widely deployed. DSL services may be
asymmetric, with upload speeds significantly lower than download speeds. Whether or not
this is a problem is largely application dependent. For example, surfing the Internet is
largely thought to have large downloads in response to a few characters of user
interaction data which is uploaded. One of us (Ellis) has noted in his own personal web
surfing that the ratio of uploaded data to downloaded data may be as high as 1:4, a
somewhat unexpected result.
Speeds are expected to be 52 Mb/sec for VDSL or (for ADSL) far less, varying with loop
length and quality. New 'modems' are also required which cost $200 or so. There's also
the question of installation, although some services (G. Lite ADSL) are user-installable.
Of course the high speed 'last-mile' service of xDSL (or any other broadband data
service) is only as good as the complete path from subscriber to ISP. Subscribers should
be aware that their path after reaching the telco switching office may include being
aggregated or multiplexed onto a circuit with many other subscribers. This might lower
their throughput and response time if not properly engineered to handle large loads.
Cable Data Services
Although many people probably do not realize it, the tree-and-branch topology of existing
cable TV networks means that these services are a shared resource so there are
concerns of lowered performance as more subscribers join. Cable operators plan to
address this, like the telcos cited above, through incremental fiberization guided by
continuous performance monitoring. Thus, the cable network's topology will slowly evolve
so that ever-fewer customers (ultimately, just one -- if traffic rises by enough) use each
cable segment. In addition, there are security and privacy concerns, which cable industry
standards will address through per-user encryption.
Speeds are also asymmetric, with faster download than upload speeds as previously
noted. Non-standard modems are required although standards are being developed. Some
early ones are 1-way, thus needing phone lines for uploading data. Also, lack of 'plug and
play' user installation has been a constraint. And since cable and PC technicians have
limited cross-training in each other's fields, they've had to be sent out in pairs, doubling
The typical requirement of connecting the cable data interface to the computer via
Ethernet requires an Ethernet interface on the computer. Typically, home computers do
not have an Ethernet interface as standard equipment and business computer users may
not want to tie up an Ethernet port this way.
Currently, services are provided on geosynchronous Direct Broadcast Satellites (DBS)
which are definitely one way because the equipment used on customer premises is receive
only. A second data service is needed for uploading. New Low and Medium Altitude Earth
Orbits (LEO/MEO) communication satellites that Iridium and other new mobile service
communications operators use, may enable two-way service, including mobile.
Non-standard 'modems', and radios are required. And the shared resource nature of the
service limits capacity in the short term.
Bandwidth and cost are the primary concerns, plus the need for new (and non-standard)
"modems". Other technical issues include the need for somewhat new technology for
computing such as radios. Novel concerns include 'line-of-sight' requirements and
possible signal blockage by rain or other environmental factors! For "wireless cable"
operators and terrestrial TV broadcasters, a further issue is their one-way radio
infrastructures, which (like cable TV) were optimized to deliver non-interactive
Digital Subscriber Line
Although Telcos, ISPs and others are beginning to widely advertise DSL services, cost and
availability are definite problems. Also the ISDN debacle continues to haunt the minds of
potential subscribers. Some have suggested (see the 1998 Harvard workshop at
http://www.ksg.harvard.edu/iip/ngct/ngct.html) that the Telcos would really like their DSL to be a
mostly unregulated, non-common carrier service.
DSL availability is also tied up in the deregulation confusion from the 1996
Telecommunications Act which sets the scene for a clash between the normally state
regulated telephone services and federal laws and regulations. Access to these services
by ISPs and Competitive Local Exchange Carriers (CLECs) continues to be a problem.
Telcos, despite recent Supreme Court decisions against them, continue to litigate against
(and thus delay) competitive service providers.
Cable Data Services
Cable services, while regulated, are not common carrier services. This means there is no
ISP choice without extra cost. Some fear that small ISPs who shelter controversial free
speech could thus be forced out of business. The issue is important enough to existing
ISPs and 'portals' that AOL is actively campaigning for these services to be 'open',
although they are not suggesting that they be assigned common carrier status. A
coalition of advocacy groups for opening cable networks to any ISP, may be found at
http://www.nogatekeepers.org. Availability also continues to be a problem, partly because cable
operators feel little competitive pressure to quickly offer broadband data service.
Some feel that the best way to get meaningful competition to the telcos is to let cable
operators 'do their thing' without requiring them to open their services to competitors.
Others feel that because the telcos are under pressure to open their systems to
competition, it is only fair to place the same requirement on cable operators.
While subscribers may always discontinue service, the use of non-standard 'modems'
means that an equipment investment inhibits transfer to another type of service. Also,
the cable companies would prefer to rent the interface device to you which decreases
competition. Possible standardization forced by the FCC is looming which would bring an
element of competition to the sales of the interface equipment. But until universal (DSL,
Cable, Satellite, etc.) interface devices (similar to the universal handsets which have been
lacking in the United States' non-standard digital cellular situation) become available, the
requirements for investment in fairly costly interface devices constrain ISP choice.
Another current policy issue is industry mergers as typified by the AT&T/TCI merger.
AT&T will get access to homes over TCI's cables and can begin to offer competitive
telephone and other services. A side issue has been the attempts of some local
governments to deny approvals for the 'new' TCI until they open their systems to
Costs and competitive issues are important. Satellite television services are about where
cable television was a decade or more ago, but the adoption curve has reached the point
of rapid increase. Satellite data services are in their infancy. The newness makes it
difficult to predict where policy issues might surface.
As usual with a new technology, esoteric issues such as interactions between FCC
antenna regulations and local regulations limiting their placement will be discovered.
Antenna placement regulations may be a particular problem for apartment and
condominium dwellers whose antenna cannot 'see' a desired satellite from their own
window, balcony, patio or similar attachment point.
Consumer terrestrial wireless data access is even less developed than satellite services.
Issues include spectrum allocation, though a solution may in time emerge from the FCC's
new debate on allowing ultra-wideband, micropower CDMA services. FCC rules (also cited
above for satellite services) that restrict outdoor line-of-sight radio antennas, are
another concern in some housing complexes.
Deployment of broadband data services is well underway. Significant, but not
insurmountable problems exist. Perhaps the biggest is that this is all caught in the
deregulation of telecommunications. For example, many industry analysts agree that
telcos' congressional lobbying before passage of the 1996 Telecommunications Act, and
their litigation afterward, have impeded competition in the markets they dominate. In
1998, MCI and others suggested that to counter this, telcos should be made to divest
their copper loops. While there is now little interest in such drastic measures, continued
failure to achieve local service choice for most customers could spur demand for them,
and even for parallel steps against cable TV operators.
A special symposium sponsored by the Harvard Information Infrastructure Project called:
'Next-Generation Communication Technologies: Lessons from ISDN'
(http://www.ksg.harvard.edu/iip/ngct/ngct.html) was held in June 1998. While the overall
results were inconclusive (everyone basically said "it wasn't our fault") there was a
serious attempt to learn the lessons of this failed three-decade effort to digitize the 'last
First, stable standards were very late, and their implementation later still ... yet for the
residential customer who must rely on ISDN for both voice and data, important features
are lacking even now. Among these are convenient extension telephones and easy
interfaces to consumers' existing analog Customer Premises Equipment (CPE) such as
phones, answerers, auxiliary 'ringers' (e.g., lights and gongs), wiring devices, etc. The
market seemed to offer windows of opportunity in the early 1980s and again in the
mid-1990s, but telcos were too slow-moving to exploit these -- and some say that
utility regulation left them little reason to do so.
Many states' ISDN tariffs also wrongly priced the service, as customers saw it. Telcos'
chronically weak grasp of users' data needs, and their poor record in order-handling and
tech support, were further obstacles to success. And AT&T's 1984 divestiture sowed
dissension among the companies that had to collaborate in order for ISDN to do well in the
U.S. (The service won greater acceptance overseas.)
DSL -- and cable modem service as well -- are immune to some of the problems that
beset ISDN, but vulnerable to others. Telcos and cable TV firms have had little success
at selling and supporting advanced data services for consumers. This suggests that ISPs,
CLECs, et al., be helped to do so. But the incumbent owners/operators of wired
infrastructure in the U.S. typically disfavor such ideas.
Experience in Canada may be useful on this point. That country's counterpart agency to
the FCC ordered that cable TV networks be opened to all data communication service
providers who wish to serve cable modem customers. Many in the U.S. will observe these
arrangements with interest.