Case Western Reserve University (CWRU) is continuing to evolve its
highly successful campuswide communication network, CWRUnet. CWRUnet was
based from the outset on the vision of what a campuswide network should
be. Today we are on the threshold of another significant chapter as we
upgrade the network from baseband to broadband technologies.
Currently, nationwide, most students who have access to campus
networks from their residence use speeds in the range of 9600 to 38.4 K
baud. A few use 64 K bits per second (bps) ISDN, and some even have 10 M
bps Ethernet or 4/16 M bps Token Ring access. Faculty and librarians are
in better shape, with higher-speed (4/10/16 M bps) networks in their
offices, labs, and classrooms. Most campus network backbones are using
10 M bps Ethernet or one of several Token Ring technologies, and many
have started to use 100 M bps FDDI (fiber-distributed data interface).
In terms of future applications, especially ones saturated with
multimedia data, these networks will prove to be inadequate. Recognizing
this, some universities are developing their own advanced prototypes of
ATM (asynchronous transfer mode)-based networks. ATM technology is
useful both for connecting end-user devices. (Fig. 1 shows a paradigm of
ATM used in a local area network, and Fig. 2 for running over network
backbones.) In some trials now being conducted, ATM is being used to
replace outmoded Ethernet in connecting servers to clients. (Fig. 3
shows one possible way to use ATM in a campus network when high
performance at selected workstations is a requirement.)
On our campus, we are using ATM to upgrade our FDDI backbone, which
is becoming overloaded. We are also upgrading linkages to several of our
servers using switched-ATM at 155 M bps speeds. (Fig. 4 shows the
resulting topology of a campus network when ATM is used to upgrade the
backbone.)
For evolving campus networks, both types of upgrades--backbone and
client-server linkages--will be needed. ATM technology offers benefits
not available with other, older technologies, such as simultaneous
voice, video, and data transmission.
20/20 Vision for a Campuswide Network
The thirty-year strategic point of view (developed in 1988) includes
elements of the networking strategy being used at CWRU.
* There is one universal and ubiquitous network for the campus: the
maximum value of the network is achieved when everyone has access to it.
This is more than being democratic; it is efficient. When only part of
the community has access, the network never can be depended on for
communicating to all, and other distribution systems have to be used,
which incurs additional expenses. In universities, it is more efficient
to use the "library model," not the telephone model, for giving the
campus community access to the network.
* All five families of communications services are supported by the
campus network: voice, video, multimedia data, telemetry, and control
signaling. Examples of services include cable TV (thirty-eight channels
are available at CWRU); videoconferencing; a campuswide information
system offering e-mail, course-specific bulletin boards, Internet
access, and the library's online catalog; surveillance for parking lots
and security sensors for campus buildings; and electronic door locks and
energy management control signals, enabling campus buildings to become
"smart."
There is only one comprehensive wiring plant on the CWRU campus;
all activities involving the transmission of information use CWRUnet. At
present, the aforementioned five families of communications services
reside on fully parallel wiring using independent optoelectronics and
software and are managed centrally. Eventually, these network-based
services will be joined in a single broadband, integrated, fiber-optic
network.
* The campus network is fast enough so that there is never a
bottleneck. The network's speed is scalable as the user's need changes.
Not everyone needs 100 M bps FDDI, but those who do should be able to
have it without having to install their own private network. At CWRU,
our fiber-optic wiring to each and every desktop provides the physical
layer for all types of communications needs, whether it be for an
electronic door lock or a highly parallel supercomputer.
Implementing the proper backbone network topology across the campus
is more difficult and depends on the needs of users. Usually, the best
tactic is to install more than sufficient capacity in the backbone's
optoelectronics, running over optical fiber cabling. Other tactics
include setting up parallel networks by replicating cabling and
electronics for the backbone and then providing several network
crossover bridges. Another possibility is to use router devices that
even out the network traffic across different cable paths.
* The network is built using a wire-once architecture independent
of electronics and software. This justifies the network as a many-year
(thirty or more years') strategic investment. We believe that campus
networks will need to be continuously upgradable and we want to ensure
that the network never inhibits use of such improved technologies. In
fact, the wiring should not need to be changed, because with our system
of wiring patch panels, different network topologies can be implemented
without reinstalling the cabling.
Since we were wiring for the long term, we needed to find cabling
that would be suitable for a significant period of time. The answer was
obvious: fiber-optic cable. But which kind: single mode or multimode? We
have used both because we saw the short-term value of multimode fiber as
premise (in-building) cabling and the strategic value of single mode,
which is capable of gigabit and, eventually, terabit transmission rates.
Even the relative cost of fiber-optic cable is not a big problem when it
is amortized over its long, useful life.
* The campus network integrates with the metropolitan area
telecommunications grid. As more and more institutions offer distance-
learning educational programs, the local, regional, national, and
international telecommunications grids become relevant. We have built
CWRUnet to be compatible with the technologies that will be used by our
regional telecommunications carriers. Our backbone signaling
technologies will match those to be used by both telephone companies and
cable TV companies. This will involve the substitution of computer
communications technologies for next-generation telephone transmission
technologies. Because our wiring is independent of the new electronics
and software systems, we expect to encounter few problems installing
those technologies as they are needed. We see the university of the
future not as insular, but as enmeshed in the telecommunications grid
for both supplying and using all information services.
* The campus network is based on standards for signaling and
protocols. Over time, we see that the computer data transmission
standards such as Ethernet, Token Ring, and FDDI give way to scalable-
speed ATM and Synchronous Optical NETwork (SONET) standards. SONET is
the ultrahigh-speed transmission technology running over fiber that will
deliver gigabit speeds (from at least 51 M bps). ATM technology uses
small, fixed-sized packets, called cells, that will carry all formats of
digitized information. Because of uniformity in cell size, ATM uses a
relatively simple technology that can rapidly switch voice, video, and
audio information without suffering the effects of distortion that other
technologies such as Ethernet can produce. This simplicity of format
allows hardware to route (i.e., switch) information packets much faster
than the software-based routing for multiprotocol data networks we use
today.
Most important, SONET and ATM are families of technologies on which
the common carrier telephone companies have agreed to standardize, so
that private campus networks that use those standards for backbone
connectivity will be able to interface efficiently and effectively with
the public common carrier networks. ATM is a scalable technology,
allowing it to handle both very high and relatively low speeds. Thus ATM
promises a simpler, scalable upgrade path as end-user needs change. Data
protocols such as Internet Protocol and IPX (from Novell) are
implemented above the ATM network layer, so that no incompatibility
exists between these familiar protocols, now being used widely on
university campuses.
* The campus network is open, permitting a wide variety of
different types of end-user equipment to be connected. No network can
accommodate all devices and interfaces, and some limitations are
necessary. Yet it will be possible, with effort, to provide some type of
translation, or gateway, for interfacing special devices.
Beyond Ethernet
The most prevalent networking technology on the CWRU campus today is
Ethernet, following the 10Base-T standard, even though we run the signal
over multimode fiber cabling to each desk. Ethernet was invented in the
mid-1970s for workstations with barely enough processing power to
execute 1 million instructions per second (MIPS). Currently, we are
using Ethernet with microcomputers capable of more than 100 MIPS. In a
few years, machines with ratings of 400-1,000 MIPS will be relatively
commonplace.
How Will Campus Networks Keep Up?
The transition from baseband networking technology, such as Ethernet, to
broadband technology will represent a fundamental change. Baseband
technology handles a single communications channel on a single wire,
whereas broadband technology supports the transmission simultaneously on
a single wire of multiple channels of information. When all information
is converted to a digital format and transmitted in ATM cells, the
cabling efficiency increases by orders of magnitude.
For most academic applications of networking, relatively large
quantities of data are transmitted infrequently, so-called bursty data,
for which packetized transmission is optimal. For time-sensitive
information, such as voice or video/audio, packetized transmission can
be problematic. A packet of digitized video produces jerky frames that
are difficult to watch. Multimedia applications form a composite of
computer data, voice, music-quality audio, and color video segments, and
transmission requires that data packets arrive at the appropriate time.
We expect that ATM, with or without SONET at the lower network layer,
will become the preferred transmission technology for campus networks.
In the future, we expect to see standard microcomputers with two or
three network interfaces--one each for data, voice, and video. Such
microcomputers will incorporate basic telephone and/or basic television
into the computer, capable of managing telephone calls and messages. If
national television standards for color or high-definition television
are included, this programmable computer can be used to provide only the
video material needed. Access to 500 TV channels will give way to video-
based information on demand. Such a network, coupled to high-capacity
information repositories, will transform the library as we know it.
Important Applications for Future Campus Networks
As the library of the future becomes extended to add software libraries,
digitized image collections, and digital books, journals, and
newspapers, the campus network will become an integral component.
Whereas the library of the past concentrated on having information
stored just-in-case, the essence of the library of the future is having
information delivered just-in-time. In the future, it will be the size
of information conduits rather than collections that will measure the
efficacy of the best libraries. This may be an oversimplification, but
it puts in perspective the importance of the campus network in the
evolution of a university's library.
By the same token, the classroom of the future is wherever the
student is. Networks will bring together students and professors without
regard to time or distance, and groups of students will use networks to
learn together.
Quo Vadis?
At the end of the twentieth century, our campuses will need to build a
new utility infrastructure for communications technologies. That new
utility will offer customized, computer-based information resources, and
it will enable the expansion of both telephone and television service
offerings.
There is a significant need for communications standards, analogous
to standards for 110-volt electrical AC. There is also the need to
direct increased efforts toward strategic planning for campus networks,
since we cannot afford to rewire our campus buildings every ten years.
If the wiring issues can be settled through the use of fiber optics and
wire-once architectures, then we can add features by updating the
software in network hubs and in central switching units.
The development of CWRUnet was made possible through a series of
intersecting visions that led to a fiber-optic network, assembled the
first elements of the CWRU Electronic Learning Environment, and
challenged faculty and students to take advantage of the new facilities.
Today the campus network is a well-used utility. Tomorrow CWRUnet will
extend the reach of the university beyond the campus in order to
facilitate learning, discovery, and evaluation in the community beyond.
Raymond K. Neff, Ph.D., is vice president for Information Services at
Case Western Reserve University.