DSL
on the Campus
an
EA White Paper
Ethernet via the telephone wire: Will ADSL work for you?
Abstract
New products have broken through the conventional limits on the use of telephone
wiring to carry high-speed data. They can be used to advantage in many settings,
such as student residences, where telephone service is normally already installed.
This article examines the suitability of ADSL for extending Ethernet to on-campus
and off-campus users, and compares it to cable modems and structured wiring
for on-campus applications.
Introduction
A technology developed for telephone companies offers some interesting possibilities
for the campus. Voice and data network organizations must work together to achieve
the benefits. The dramatic new capabilities recently brought to the market in
Asymmetric Digital Subscriber Loop (ADSL) products make it possible to deliver
high-speed data service over the same wire pair already used for voice. ADSL
and its newer, lower-speed variant called G.Lite or DSL.Lite, are members of
family of similar technologies that are collectively referred to as xDSL. In
the right circumstances ADSL and G.Lite can be very attractive alternatives
to new copper and fiber data cabling as a cost-effective ways to extend Ethernet,
but only Ethernet, to locations that are not yet on the campus data network.
On-campus
Many organizations have already used earlier versions of DSL technology for
low-cost extension of Ethernet to low user-density buildings by using spare
pairs in voice cables. This is still a practical approach where there are few
users to be served, making wiring inside the building a minor expense. The distinctive
characteristic of ADSL (and G.Lite, which will not be mentioned explicitly again,
but is implied wherever ADSL is referred to in this document) is that it operates
on the same wires that are already in place for voice telephone service. On-campuses,
with their own voice switches (PBXes), this can be a major advantage. For campuses
using a telephone company provided voice switching (Centrex), it might be a
significant complication. We will start with the PBX case and come back to Centrex
later.
Using the existing voice pair makes ADSL especially well suited for bringing
network service to dormitory rooms and student apartments. The very fact that
there is telephone service in place to each residence space means that the necessary
physical wiring has been done. All that needs to be added to implement ADSL
are the desktop electronics in the living quarters and a rack of equipment at
an upstream location where the copper pairs from these living spaces are all
accessible.
The equipment to be installed in the dormitory room or apartment is a cigar
box sized modem. An internal or external filter, usually referred to as a "splitter",
is used to separate the voice and data signals on the wire. The splitter is
plugged in to the phone jack, the telephone and the ADSL modem are plugged in
to the splitter, and the computer is connected to the data port on the modem.
(Figure 1) G.Lite is supposed to make the splitter unnecessary. However early
field experiences suggest similar devices (micro-filters) installed on each
voice phone connection, are often needed. Some modems come with the splitters
built in so that no separate device is necessary.
The ADSL modem's function is similar to that of the familiar dial-up modem,
putting a data signal into a form that can be transmitted across a copper pair,
but the ADSL modem has several important differences:
-- It does not interfere with the use of the line for regular voice traffic. The user can browse the Internet and talk on the phone at the same time.
-- ADSL works at much higher speeds than a dial-up modem, giving downstream data rates (from the network toward the user) of up to 8 megabits per second (Mbps) over the copper pair, compared to the 56 kilobits per second (Kbps) possible with dial-up connections. The upstream data rate for ADSL can be as high as 1 Mbps. G.Lite works at speed up to 1 Mbps downstream and 384 Kbps upstream. Both ADSL and G.Lite operate at lower than their peak speeds on phone lines that are long or have certain other impairments, but these impairments are not likely to be a factor in a campus setting. (An emerging version of DSL, Very high-speed DSL (VDSL), is expected to offer downstream speeds of 26 to 52 Mbps on campus-scale voice cable plant in 1999 or 2000.)
-- It is always "On". Whenever the user's computer is "On" it is connected to the network, with none of the dialing and log-in delays that discourage dial-up use.
-- It connects to the Ethernet port rather than to the serial port of the user's computer. To the computer it looks like any other Ethernet network device. The ADSL modem buffers the traffic so that the difference between the actual line speed and that of the Ethernet port is essentially invisible to the user.
-- The data connection is effectively a dedicated line and is made without using up any of the control or switching capacity of the PBX. The use of dial-up modems for prolonged connections has caused switch capacity problems for many voice network managers in recent years.
-- The dedicated line nature of the data connection improves network access security by eliminating the need for ADSL users to connect to the network over dial-up connections. The number of dial-in modems can be reduced, cutting down on the exposure to dial-in attacks An unauthorized party would have to physically tap into an authorized ADSL user's line in order to pretend to be the authorized user, because there is a correlation between physical ports and users.
At the other end of the copper pair there is corresponding equipment that separates
the high-speed data from the voice and multiplexes the data from many users
into a single Ethernet, Frame Relay, or ATM data stream for connection to the
rest of the network. The arrangement is commonly known as a DSL Access Multiplexer
or DSLAM.
The most obvious location for the DSLAM is in the campus wire center. On many campuses all the voice wiring is on copper pairs that run from the phones to the wire center. In these cases one large DSLAM can serve many buildings and there is an economy-of-scale advantage to be enjoyed. On those campuses where some of the voice lines are served out of equipment cabinets in other locations, DSLAMs can be co-located with that remote voice equipment. The DSLAM can also be still closer to the user, i.e. at some intermediate point on the copper pair, such as in the basement of a high-rise building, if that is desirable. When DSLAMs are distributed around the campus, the upstream connection (uplink) for the aggregated data traffic might be made at each DSLAM location. (Figure 2) Alternatively, each uplink might be extended to some other data network access node, presumably over the same type of high-speed copper or fiber technology that links the voice remote systems to the campus wire center.
Figure 1: ADSL in PBX Room
Unfortunately, ADSL is less certain to be a cost-effective way to bring Ethernet to offices and classrooms. Where these spaces have single-line, analog phones, it will work just as it does in dormitory rooms. However, many classrooms do not have telephones at all, so there is no existing wiring to exploit, and many offices have multi-line phones that make it more difficult to connect the modem to a particular pair. Also, some offices have phones using digital technology that is incompatible with most of the currently available ADSL equipment. This last problem is fairly straightforward to overcome, and a few vendors already offer solutions.
Figure 2: ADSL Remote
Off-campus
While the short cable lengths of the private telephone cable plant on the campus
make it an ideal setting for ADSL, the public telephone network for which the
technology was developed contains a number of potential technical and regulatory
impediments to its use. These make it more difficult, and often impossible,
to use ADSL as the solution for off-campus users who desire higher speed connections
to the campus network.
If the telephone company (Incumbent Local Exchange Carrier, ILEC) or a competitor
(Competitive Local Exchange Carrier, CLEC) offers ADSL services to its customers,
it is theoretically possible for at least some off-campus locations to be equipped
for the service and to use it to access the campus network. To get from theory
to reality, two technical factors have to be satisfied:
-- The ILEC or CLEC must have equipment that permits different users to be connected
to different destinations, in this case to the university network instead of
just one particular commercial Internet service provider (ISP). The ability
to do this adds complexity and cost to the ADSL network, but many carriers are
offering it so that more than one ISP or employer of telecommuters can use those
carriers' networks to offer high-speed access.
-- Each user has to be served out of a wire center that is equipped with a DSLAM
and on a phone line that is technically capable of supporting the technology.
A surprisingly large proportion of the public network lines that are adequate
for voice service will not support ADSL, sometimes for insurmountable technical
reasons and sometimes due to characteristics that could be overcome. Some telephone
companies have shown a remarkable unwillingness to do anything about deficient
lines that could be made serviceable. There is no regulation that requires service
providers to make ADSL available to anyone, even in cases where it works for
one customer but not the next door neighbor.
Then, of course, the price for the service has to be tolerable. Neither the
technical nor the cost conditions can be taken for granted.
The Telecommunications Act of 1996 does require the ILECs to make spare copper
pairs available to CLECs, and a university could go through the regulatory process
to attain CLEC status in order to build a private network on telephone company
wire. Even then, practical problems associated with establishing a private ADSL
operation (DSLAMs in all the relevant wire centers, the phone company having
spare pairs where needed, users not too far from the DSLAM, etc.) could easily
outweigh the possible benefits for anyone not intending to offer the service
to the public.
Centrex Considerations
There really are not any technical barriers here, but at the time of this writing
(January 1999), attempting to implement private ADSL on a campus with Centrex
voice service could involve breaking new regulatory ground. It will center on
who owns the wire and, if the university owns the wiring on the campus, on who
has the right to decide whether ADSL can be added to lines carrying Centrex
voice service. So far, the ILECs have been successful in refusing to let CLECs
add ADSL to ILEC lines already used for ILEC provided voice service, even though
that is how ILECs use the technology themselves. The ILECs will be tempted to
take a similar position with regard to lines carrying Centrex service, if only
to be consistent with their resistance to the CLEC requests. This aspect of
making ILEC "network elements" available to CLECs is not specifically
addressed in the Telecommunications Act, and at least one court has made a decision
favoring the ILECs.
The Centrex provider might counter-propose furnishing ADSL service of its own
to these on-campus user locations. The need for them to make a profit on it
is sure to make it more expensive than a private implementation over the long
run, but a low recurring cost rather than a large capital expense might have
its own appeal. In order for this to work, however, the phone company has to
be equipped to deliver the traffic from the campus users to the campus network.
This might consist of what amounts to a dedicated ADSL installation on the campus
by the phone company, or it might involve the "different users to different
destinations" capability that applied for off-campus users in the preceding
section.
Alternatives
As already mentioned, there are other DSL technologies that can carry relatively
high-speed data over campus scale lengths of voice grade copper pairs. These
lack ADSL's ability to operate on the same pair already in use for voice, so
using them would involve at least adding new drops from the telecom closet to
the user location.
Cable modems are another significant option in locations served by a university-owned
CATV system. The modems being used in the CATV industry today are very different
from their ancestors, the 10Broad-36 units used on private broadband cable systems
on-campuses several years ago. The new units handle both directions of transmission
on a single coaxial cable, rather than separate cables for each direction of
transmission. The key drawback here is that most commercial and private CATV
plant was designed and installed to provide one-way transmission only. The new
modems cannot work in the intended manner unless the CATV plant has been upgraded
to support two-way operation. There are cable modems that get around this by
using a conventional dial-up modem connection over the telephone network for
upstream traffic. This is very slow compared to the upstream path that ADSL
provides, and it continues to burden the voice switch with protracted connections
for data traffic. Upgrading CATV cable plant to support two-way traffic is an
expensive proposition, but where it has been done cable modems can provide high-speed
connections independent of the telephone network.
Users of cable modems on public networks have legitimate concerns about privacy,
which vendors are now addressing with encryption, and about performance, which
can be seriously degraded if the service provider tries to put too many users
on a single segment of the network. In a campus setting these concerns are much
less likely to be applicable because people are accustomed to being on a shared
network and the user community is less likely to be big enough to create serious
congestion problems.
Costs
The cost of a private ADSL installation could be anywhere in a wide range, depending
on the scale of the installation and the capabilities included. The list price
to provide minimal capabilities to 1000 users out of a single DSLAM location
currently is in the neighborhood of $1000 per user. Before you let that deter
you, bear in mind that some vendors have been known to offer discounts in the
50% range on proposals for actual jobs. Also, these prices are likely to fall
as manufacturers recover their development costs and as production volumes increase.
This price does not include any changes to the existing data network to accommodate
the added user load. Since that will probably be just one direct connection
from the DSL concentrator to the campus backbone, the incremental cost per user
should be negligible. The major offsetting advantage over alternatives here
is that nothing new is needed to carry the data across the campus.
There are proprietary variations of DSL that are less expensive, but these will
not be interoperable with the G.Lite modems that some computer manufacturers
are planning to start building in to their products soon.
A cable modem solution for this same user community would have a list price
of about $350 per user. That is on top of the cost of upgrading the CATV plant
for two-way operation. The cost of a two-way upgrade is too specific to the
particular network to estimate here, but that upgrade would have to be very
expensive before it made the cable modem option more expensive than ADSL. This
does suggest that anyone considering ADSL probably ought to also investigate
the suitability of cable modems for the job.
These should be compared to an actual cost of about $350 per user for professionally
engineered and contractor-installed structured cable in a residence hall setting,
including modest but adequate Ethernet electronics. This approach will also
require high-speed data connections from the residence building(s) to a campus
data network access point, if that is not already in place. As with the cable
modem case, the cost of that cross-campus connection is too location specific
to estimate here, but it can be pretty expensive and still compete well with
ADSL.
Conclusion
ADSL warrants serious consideration as a way to provide high-speed data network
access where it has not been practical before, especially in student residences
on campus and other settings with analog phone service. It can be installed
quickly and with minimal disruption. Campuses that use Centrex for voice service
might find themselves enmeshed in regulatory issues, but it is worth at least
exploring the telephone company position on the subject. Its potential for off-campus
use is heavily dependent on the attitude of the ADSL service provider with regard
to where such service is offered, what effort they will make to condition lines
to make it work, and to what destinations users of the service are able to connect.
Whether it can be used for on-campus or off-campus users, the peripheral benefits
will include reduced load on the voice switch and improved network security
from the reduced use of dial-up modems.
However, ADSL is relatively expensive, at least for now. Cable modems might
be a less costly choice if the performance they offer is adequate, and if a
two-way CATV plant is in place or the current plant can be upgraded.
Where you can envision a need for higher speeds, and where there are no major
barriers to construction, structured wiring is still the way to go. Where new
construction is not practical, network planners might want to wait for an emerging
variation on the DSL technology, called Very high-speed DSL (VDSL), which is
expected to deliver downstream speeds of up to 52 Mbps on campus-scale telephone
networks by late 1999.