Ref: 99980049
Title: Internetworking
Date: 11/28/88

Copyright 3Com Corporation, 1991.  All rights reserved.

.h1;Introduction

Computer networks are very much a part of today's workplace.
Local area networks (LANs) or wide area networks are used
extensively to solve data communications problems.  They allow a
variety of computing devices ranging from personal computers and
workstations to minicomputers and mainframes to communicate.

As organizations grow, so do their computing and networking
needs.  Departments swell, divisions mushroom, and branch offices
open in various locations around a city or across the country.
It is common for one organization to have many networks, widely
ranging in size and performing different functions.  In an
office, a PC network enables users to share files and printers.
In the engineering division, network connections provide
workstation-to-host access for computer-aided design
applications.  Elsewhere, a remote manufacturing site's network
suports computer-aided manufacturing functions.

With the wide acceptance of computer networks, a new requirement
is clearly evident - internetworking.  Internetworking, quite,
simply, is the ability to connect multiple networks.  The most
obvious benefit to creating an internetworking system is
expanding the network size.  Users on one network can access
resources on another network.  Communications among divisions or
remote locations can increase dramatically.  Other benefits are
not as apparent.  For example, segregating networks can ease
management and directing traffic flow can achieve higher data
throughput.

Once linked, the various networks become subnetworks to the newly
created larger network called the internetwork.  The internetwork
is the communications foundation for the entire organization.  It
provides the computing connectivity necessary to ensure high
efficiency and productivity.

Internetworking products enable interconnection of both similar
and dissimilar networks.  Interconnection of dissimilar networks
brings up a particularly complex set of issues.  Networks use a
variety of communications protocols such as XNS, TCP/IP, or OSI,
as well as different technologies such as token ring, Ethernet,
and the forthcoming FDDI.  In addition, the same technology can
support different media.  For instance, Ethernet can use thin
coaxial cable, thick coaxial cable (baseband or broadband),
twisted-pair wiring, or fiber-optic cable.  In spite of the
difficulties arising from this wide variety of technologies,
however, solutions to these connectivity problems do exist.

An array of internetworking products - repeaters, bridges,
routers, and gateways - is currently available to network
designers.  Each serves a different purpose.  To select the
appropriate internetworking product requires a thorough
understanding of the application.  There continues to be much
confusion as to the benefits of and difference among these
products, especially in the case of bridges and routers.

The primary purpose of this tutorial is to clarify the functions
and appropriate applications of each internetworking product
type.  After a brief review of the four major product
classifications, there is a more detailed discussion of two
internetworking products - bridges and routers.  These products
require further examination due to their rising popularity and
the confusion that surrounds their respective application.
Descriptions of their unique benefits, the problems they best
solve, and how they can coexist on the same internetwork are all
discussed.  To provide greater understanding, there is also an
explanation of how these devices work.  Finally, this tutorial
offers guidelines for network managers to consider when choosing
among these devices for particular applications.

.h1;How To Use This Tutorial

The major sections of this article can be read consecutively or
as independent units depending on the level and type of
information needed.  For a more complete view of internetworking,
read the entire guide.  For a brief summary, read the
Internetworking Solutions section.  Should you encounter any
unfamiliary terms, refer to the glossary.

.h1;Internetworking Solutions

Internetworking products fall into four broad categories -
repeaters, bridges, routers, and gateway.  Each handles different
functions that directly correspond to the ISO layer at which the
internetworking function is performed.  To understand these
functions, it is necessary to be familiar with the role of each
layer in the ISO model.  The first diagram briefly describes this
model.  The next diagram illustrates the ISO layer at which each
internetworking product operates.

                      ISO Reference Model
------------------------------------------------------------------
                    Open System Interconnect
    Layer                           Function

7  Applicaton       Specialized functions such as file transfer,
                    virtual terminal, electronic mail

6  Presentation     Data formatting and character code conversion

5  Session          Negotiation and establishment of a connection
                    with another node

4  Transport        Provision for end-to-end delivery

3  Network          Routing of packets of information across
                    multiple networks

2  Data Link        Transfer of units of information, frames, and
                    error checking

1  Physical         Transmission of raw data over a
                    communications channel


            Internetworking at Multiple Levels
------------------------------------------------------------------
    7      Application   <---------
    6      Presentation           |   Gateway
    5      Session       <---------
    4      Transport
    3      Network       <---------   Router
    2      Data Line     <---------   Bridge
    1      Physical      <---------   Repeater

Repeaters

Repeaters extend the geographic coverage of a local area network
by interconnecting multiple segments.  For example, the Ethernet
standard specifies a maximum length of 500 meters for a single
segment; but with repeaters interconnective five segments, an
Ethernet network can reach a maximum distance of up to 2,500
meters.  Repeaters can also interconnect segments using different
physical media such as thick Ethernet, thin Ethernet, or fiber-
optic cbles.  Repeaters are hardware devices that operate at the
Physical layer of the ISO model repeating all electrical signals
from one segment to the other.  They do not provide any type of
traffic isolation.

Bridges

Bridges interconnect local or remote networks at the media access
(or MAC) sublayer of the Data Link layers of the ISO model.
Bridges are transparent to high-level protocols such as XNS, OSI,
or TCP/IP.  Their main purpose is to partition traffic on each of
the interconnect segments.  Bridges forward only traffic
addressed to the other subnetworks, increasing the effective
throughput of the entire network.  Although all segments
interconnected by a bridge form a single logical network, they
are elctrically isolated from one another.

Local bridges may connect similar networks such as Ethernet or
broadband or dissimilar networks such as Ethernet to broadband.
Remote bridges use data communications links such a T-1 lines to
join physically isolated networks.

Routers

Routers perform packet routing and forwarding functions.
Operating at the Network layer of the ISO model, they link
networks that share the same network layer.  Most routers on the
market today interconnect networks that use the same high-level
protocol suites.  But there are routers, recently introduced,
that work in multiprotocol environments.  Due to their recent
arrival and their support of mixed protocols, these routers are
respectively less mature and more complex than those which
function in single-protocol applications.

Routers create a number of logical subnetworks, allowing large
internetworks to be organized into different administrative
domains.  Routers, for example, may interconnect Ethernet
segments either locally or remotely over point-to-point lines.
Another example is an applicaton where routers interconnect token
ring and Ethernet networks over an X.25 wide area network.

Gateways

The most complex of internetwork products, gateway interconnect
networks that have totally different communications
architectures.  Since the network facilities and addressing
schemes are incompatible, the gateway must provide complete
conversion from one protocol stack to the other without altering
the data that needs to be transmitted.  Currently available
gateways make such interconnections as TCP/IP to SNA or TCP/IP to
X.25.

A TCP/IP-to-SNA gateway gives users on a multivendor TCP/IP local
area network access to IBM hosts through the SNA protocols.
Gateways between local area networks and X.25 wide area networks
either connect LAN users to X.25 hosts or large databases on an
X.25 Public Data Network (PDN) or they connect users attached to
a Packet Assembler Disassembluer (PAD) on the PDN to LAN devices.

.h1;Bridges

The relatively new concept of MAC-layer bridging offers powerful
solutions for a number of internetworking problems.  Since a
bridge operates at the Data Link layer, a low level of the ISO
model, it can transparently pass traffic running different high-
level protocols.  Bridges are thus a flexible and cost-effective
choice for heterogeneous network environments.  A bridge can
interconnect networks running a wide variety of applications
using such protocols as XNS, TCP/IP, or OSI, though only devices
running the same protocols can communicate with one another.  For
example, traffic from a TCP/IP device is understood only by other
devices running TCP/IP.

In addition to interconnecting networks running different
protocols, bridges can connect networks using different media
such as coaxial cable (baseband or broadband), fiber-optic cable,
or twisted-pair.  For instance, an internetwork might consist of
several coaxial cable networks or a mixture of coaxial, twisted-
pair, and fiber-optic cable networks.  The interconnected
networks can also use different access methods, for example, CSMA/
CD or token passing.  One of the limitations on such
internetworks is the possible difference in maximum frame size
supported by the various subnetworks.  It is essential that the
high-level protocols passing through the bridge do not violate
the maximum packet size on any network segment.

Along with protocol transparency, another important benefits of a
bridge is to provide simple traffic isolation between physical
segments or cabling systems that make up the internetwork.  The
bridge accomplishes this with a learning algorithm that alerts
the bridge as to which packets should stay on the segment and
which should be forwarded to another segment.  These steps in the
traffic isolation process are known as learning, filtering, and
forwarding.

There are two main categories or bridges: local and remote.  A
local bridge interconnect two or more directly attached local
networks.  A remote bridge connected multiple physically isolated
networks by means of long-haul data communications links.
Typically, these are point-to-point links with speeds ranging
from 9600 bps to T-1.

Basic Principles

For the learning algorithm to function, the bridge listens to all
traffic on the attached segments.  It then checks the source
addresses of all packets and the location of their sending
station.  The bridge progressively organizes these addresses into
a table that it uses to determine whether a packet should be
discarded (filtered) or forwarded.  If a packet is forwarded, the
bridge consults its address table to select the appropriate
destination for the packet.  Filtering and forwarding are
relatively simple in the case of a local bridge connecting only
two networks, but become increasingly complex for local or remote
bridge interconnecting multiple networks.

As does any other node on the local segment, the bridge
regenerates each packet it receives.  Therefore, the number of
nodes on the segment or the distance the packet travels before
reaching the bridge has no effect on the quality of packets being
forwarded to another segment.  Any delay related to distance,
however, must meet the requirements of the high-level protocols
involved.

The forwarding and learning process assumes that the topology of
the overall network is a tree or that there is only one path
between any two nodes located on LANs separated by bridges.  If
active loops (parallel paths) exist, problems may occur, such as
packets being duplicated or traveling endlessly throughout the
internetwork.

In order to deal with this problem, some bridges implement
intelligent algorithms to detect loops and shut down alternate
paths.  The Spanning Tree Algorithm is one example.  If the
active path fails, one of the inactive paths takes over
automatically.

For remote bridges, parallel lines do not constitute a loop, so
bridges can balance the internet traffic among multiple lines.
This allows planners to design networks with some level of
redundancy.

Topology

The most common topologies for bridges are cascaded networks or
backbone networks for local applications and star topologies for
remote applications.  A choice among these depends on the number
of computing devices networked and how much partitioning they
require.  In a case where the performance of the network is no
longer satisfactory due to traffic bottlenecks, bridges can
divide the network into segments - forming a cascaded network.
The bridges control and monitor inter-segment traffic, restoring
the efficiency of each segment.  For example, a bridge might
isolate a group of PCs or workstations sharing the same file
server.  Whatever the case, a cascaded network should probably
include no more than five or six segments.  Otherwise, the delays
introduced by successive bridges may become excessive for the
higher-level protocols, as well as intolerable to users.

A high-speed backbone is a reasonable alternative in cases where
many segments need to be linked.  One prime advantage is that
such a configuration allows systematic network growth. In
contrast, many cascaded networks are the result of pressure from
unplanned growth.  Another benefit of backbone configurations is
improved performance, since inter-segment traffic only passes
over one intervening segment between the source and destination
segments (unlike a cascaded network where traffic must traverse
all intervening networks).

A backbone topology is extremely efficient in an office tower
with many floors.  In this case, an Ethernet backbone - either
coaxial or fiber-optic cable - runs the full height of the
building.  Ethernet ribs extend from the backbone onto each
floor.  Bridges partition the traffic among the floors,
maximizing the performance of each segment.

The star topology is the most common choice for remote
applications.  It allows remote sites to be interconnected with a
minimum number of intervening segments and without loops.  For
instance, a large corporation with several divisions and remote
sales offices can solve its connectivity problems with bridges
in a star configuration.  Divisions with high network bandwidth
requirements are connected to the headquarters through high-speed
T1 links and a 56 Kbps back-up link, whereas smaller remote
offices are interconnected with lower speed lines.

Network Management Given the complexity of an internetwork and
its many components, it is important to manage these resources
for most efficient use.  It is also critical to manage these
resources from a central location.  In this area, bridges are
privileged devices since they see all traffic on each attached
segment.  Some intelligent bridges can collect information and
display it on an attached console terminal or they can send it to
a central network management station for further analysis.

A strong network management scheme enables a network manager to
display bridge-supplied status information and statistics on a
global or per node basis.  Reports may include network
utilization as seen in the accompanying figure, number of frames
transmitted or received, collisions, or alignment errors.  This
information helps the network manager troubleshoot the network
and make future decisions regarding the network design.  For
example, if network utilization levels reach 50 or 60 percent,
the manager might consider adding another bridge or changing the
topology from cascaded to backbone.

Security and Protection

Intelligent bridges can also help the network manager control
internetwork security.  With special filters, stations or network
segments can be individually protected from specific stations or
packet types.  This type of filtering is different from the
learning filtering described earlier in that it allows the
network manager to prevent certain kinds of packets from crossing
the bridge.

The network manager can do this by specifying a particular
pattern anywhere within an Ethernet frame.  This pattern can
correspond to a protocol type, a protocol header, a source
address, a destination address, or any traffic pattern.  For
example, a bridge can filter all XNS traffic or can prevent
network nodes on a segment from accessing a large computer
located on another segment.

Performance

Another important area requiring careful study is that of
performance.  A bridge filters all the packets transmitted on
each of the attached networks.  In practical terms, this means
that the bridge must be able to receive and check packets at a
rate corresponding to the maximum anticipated usage for each
network segment.  If the bridge cannot accommodate this traffic
load, it will lose packets, causing the end stations to
retransmit them  This results in performance degradation and
possibly session disconnections.

The maximum possible usage of a network is often much less than
its theoretical limit.  A 10 Mpbs Ethernet network can
theoretically carry small-size (64 byte) packets at a rate of
14,880 packets per second.  Because of the access method used by
Ethernet, 100 percent efficiency can only be achieved by a single
station of synchronized transmitting stations.  Otherwise, too
many collisions occur and overall network throughput is
negatively affected.  Even very large Ethernet networks rarely
exceed 50 percent utilization for a long period of time.  In
practice, a local bridge that can handle an aggregate filtering
rate of 19,000 packets per second is appropriate for virtually
all traffic conditions.

Another parameter that bears examination is the forwarding rate.
Bridges partition the load on the different segments so that the
amount of inter-segment traffic is relatively limited.  For
instance, users and their file server should be on the same
network segment.  Aggregate rates of about 6,000 packets per
second are sufficient for all applications, and are within the
range of existing local bridges.

.h1;Routers

Routers are devices that interconnect multiple networks,
primarily, running the same high-level protocols.  They operate
at the Network layer of the ISO model.  With more software
intelligence than bridges, routers are well-suited for complex
environments or large internetworks.  In particular, they support
redundant paths and allow logical separation of network segments
(so each segment can have its own network number).  Routers also
better solve the problems associated with interconnected network
segments using different media such a token ring and Ethernet.
For instance, the packet size is controlled by the Network layer
and is identical on both sides.  The problem of address
resolution is also addressed in a better way in that routers,
unlike bridges, do not pass MAC layer addresses from end-to-end,
but rather each router knows the MAC layer address of the next
router in the path.  This approach avoids the delicate problem of
converting MAC layer addresses from one format to the other
before transmitting a packet.

Since routers impose no topology constraints, they provide
sophisticated routing or flow control as well as traffic
isolation.  How routers perform these functions depends largely
on the network protocol they use and the particular
implementation thereof.  Unlike bridges, routers do not require
full participation of sending and receiving stations in
addressing packets.

Routers give network managers the ability to define boundaries
for administrative control.  Using a hierarchical addressing
scheme, a network manager can divide a large internetwork into
small administrative domains.  A good example is a university
campus with a backbone network linking different departments.
With routers, each department's network can remain logically
separate and under the administrative control of the department.

Basic Principles

In contrast with a bridge which makes a simple forward or discard
decision, a router is able to choose the best route for each
packet.  It can do this because the address of the final
destination network is attached to the packet.  Based on this
information, it looks in a routing table to disconver the best
path.  In this process, it is important to understand the
fundamental difference between a bridge and a router: as it was
described in a previous section, a bridge looks at all the
packets sent on its attached network, whereas a router receives
only the packets that are addressed to it directly by either an
end station or another router.  Each node on the path knows how
to go to the next hop in order to get a certain network.  The
routing process is handled on a step-by-step basis.

Because routers selectively forward packets, loops are allowed in
the internetwork topology.  In addition, most routers implement a
time-to-live program for packets.  This process, which consists
of destroying packets that have traveled too long or through too
many routers, prevents bad packets from congesting the network.

Routing Mechanisms

The information used by routers to direct packets is updated
either statically or dynamically.  With static routing, the
network manager configures specific paths to the different
segments in the internet.  With dynamic routing, the router
itself creates and updates paths as changes occur on the
internet.  The router does this by constantly monitoring internet
activity.  Any time an important event occurs, for example, a
station is removed or added, the router revises internet paths to
reflect such a change.  Dynamic routers usually also have static-
routing capabilities.  In a pure static-routing applicaiton,
direct route control is up to the network manager.  For this
reason, it is not well-suited for large internetworks where the
topology is subject to change.

Dynamic routing relies on specific protocols to convey the
routing information throughout the internetwork.  Using RIP
(Routing Information Protocol), a component of both the XNS and
TCP/IP protocol suites, routers can broadcast their own routing
information.  RIP makes routing decisions using an algorithm
based on the number of hops between two networks.  As a result,
the router ensures that traffic travels over the shortest
possible path.  To assist in this process, the network manager
can manually change the number of hops associated with specific
routes to modify the preference for one route over another.

Topology

Routers are often the foundation of large internetworks.  They
impose no constraints on network topology and they provide a way
to divide the internetwork into domains or subnetworks.  Since
the maximum size of an internetwork often depends on the time
required for a packet to go from one end to the other, routers'
ability to select the shortest paths allows larger
configurations.  Routers also permit systematic development of
complex networks as an oranization grows.

.br;Load Sharing
By spreading the traffic load over the internetwork, routers
minimize the possibility of congestion at a single point.  A
remote router with parallel links can choose at will whichever
link is currently providing the most rapid packet delivery.
Another way to provide load balancing is to statically assign
groups of stations to different routers, preventing traffic from
focusing on a single router.  This reduces delay in the routers
and ultimately improves network throughput.

Congestion Control

Internetworks sometimes face congestion problems similar to those
on highways in densely populated areas.  This situation usually
arises from processing limitations of internetwork products or
from speed mismatches between LANs or the longhaul links that
interconnect them.  Routers can handle these problems in
different ways.

In the rare case of traffic overload, routers can drop packets
and then cause them to be regenerated later.  This is consistent
with the "best effort" approach in datagram networks.  Such a
method is often counterproductive, however, since it does not
solve the funamental problem of sources sending more packets than
the routers cna realistically handle.  To overcome this problem,
some more sophisticated routers can send special packets (i.e.,
ICMP source quench messages in the IP protocol) to the sources
asking them to slow down.  Upon receiving such a quench message,
the source learns the destination is congested so it slows or
even stops traffic to that node.

Network Management

To provide the best possible service, routers need good network
management capabilities.  For example, a network manager trying
to identify communications problems between two end nodes must be
able to trace the exact paths packets are following.  In another
case, to take advantage of a new interconnection the manager may
want to change routes in the router or even modify the preference
levels among routes.  Whatever the situation, accurate statistics
are essential for measuring and maintaining the quality of the
interconnection service.

Other prime requirements are an ability to perform all management
functions from a central location and an audit trail that
automatically collects statistics.  With these features, the
network manager can monitor traffic and then anticipate and
control any problem.  Currently available routers offer most or
all of these network management capabilities, enhancing their
attractiveness as internetworking solutions.

.h1;Bridges vs. Routers:Selection Guidelines

Network planners must completely understand the computing and
networking requirements of their environments before they can
design an efficient internetwork system.  In particular, the
choice between a bridge or a router depends on a thorough
analysis of the application.  To simplify this decision-making
process, the following section compares the relative strengths
and weaknesses of bridges and routers in specific area.

Computing Environment

In environments with a variety of computer resources (PCs,
workstations, mainframes) there are varying requirements for
network bandwidth.  For instance, clustered minicomputers,
diskless workstations, or PCs sharing a file server place a great
burden on a network because of numerous file transfers.  When
such traffic begins to seriously affect performance, the most
efficient way to deal with the problem is to use a bridge to
subdivide the network into two segments, thereby partitioning the
traffic.

Networking Environment

In multiprotocol environments, bridges provide, at this time, a
more flexible and mature solution.  They are transparent to high-
level communications protocols and accommodate many different
applications.  For instance, bridges would be a logical solution
in a case where a user wants to interconnect networks supporting
a mix of protocols such as DECnet, XNS, and TCP/IP.

Over the past few years, an increasing number of computer and
networking vendors have introduced products using the TCP/IP
protocol suite, making it somewhat of a de facto standard.
Therefore, multivendor, single-protocol environments are more and
more common.  In these cases, routers are a possible solution.
Network complexity will probably determine whether a bridge or
router is used.  In simple configurations, bridges are a sensible
choice.  As complexity increases, so does the need for the
traffic isoltion and control capabilities or routers.  A
combination of bridges and routers, though, can solve
particularly complex internetworking problems.  Bridges are used
to isolate the traffic between the different departments within
each site.  Routers are used to interconnect remote locations to
allow each site to have a separate logical network.

Network Topology

Bridges impose some restrictions on network topology.  Active
loops cannot exist on a bridged internetwork.  Routers, however,
support all network topologies.  Here the size of the proposed
internetwork and the level of redundancy required is extremely
important.

The majority of simple cascaded networks function well with
bridges.  As more networks and alternate paths are added, though,
routers become the better choice.  This is particularly apparent
in cases where there is a need for redundant data paths and load
sharing.  Router keep up-to-date on the changes in the topology
and are able to determine the shortest path between two points.

Network Administration

Bridges and routers offer equal capabilities in the areas of
network statistics and monitoring.  Both can provide automatic
audit trail and, along with a network management station,
sophisticated analysis of such information.  How an organization
wishes to administer its internetwork will most likely determine
what internetworking product it uses.

On one hand, if the network manager wishes to administer the
entire internetwork from a central location, a bridge makes good
sense.  A bridged internetwork acts as a single logical network.
For example, the network manager must take into account all
stations on a bridged internetwork when adding or removing a
station.

On the other hand, an organization may want or need decentralized
network management, thus making a router the better choice.  An
internetwork connect by routers allows each segment to be
logically independent.  Internetworks that have many distant
sites or that are relatively large may require several network
managers.  For instance, each department in a university may want
to retain control of its own network while still retaining access
to other networks on the same or other campuses.

For ease of installation and maintenance, bridges offer definite
advantages over routers.  Bridges require no intervention from
the network manager.  They can make extremely basic routing
decisions themselves.  Routers, however, are more sophisticated
devices.  Particularly if static routing is used, they require
network manager intervention to establish and maintain the
desired configuration.

Performance

Because the operate at a low level of the ISO model and perform
relatively simple software tasks, bridges provide higher
throughput than routers.  Routers, though, offset their slower
performance by performing sophisticated software tasks such as
dynamic routing.

.h1;Conclusion

It is important to remember that each computing and networking
environment has its own characteristics.  In some cases, the
decision regarding which internetworkings solution to implement
is easy.  In others, the choice is less obvious.  Those
responsible for network planning must carefully consider current
computing resources and how to best internetwork them.  There
must also be thought for the future.  How is an organization
likely to expand and what kind of internetwork could best support
it?

For example, do your networks run one or many different
protocols?  Bridges, currently, have the distinct advantage over
routers in multiprotocol applications.  What sort of network
management is best for your organization?  Routers allow internet
segments to act as individual networks, while bridges function as
a single logical network.  Routers allow internet segments to act
as individual networks permitting independent administration of
these segments.  Does your internet have two or three segments or
does it have twenty?  Bridges efficiently interconnect a
relatively small number of networks, whereas routers can capably
handle more complex topologies.

Once this kind of analysis is complete, one of the other of these
internetworking products will emerge as the appropriate choice
for the particular application.  In some cases, the best
configuration might include both bridges and routers, or you may
start out with bridges and add routers as your network expands.
The implications of this concept are particularly important,
especially in relation to growth.  In any event, no matter what
internetworking solutions you choose today, your future options
are unlimited and your current investment is protected.

.h1;Glossary
Access Method       Way to determine which workstation or PC will
                    be the next to use the LAN.  A set of rules
                    used by network software and hardware that
                    direct traffic over the network.  Examples of
                    access methods are token passing and
                    collision detection.

Address             A set of numbers identifying the location of
                    a node on the network.  Each node must have a
                    unique address on that network.

Audit Trail         A function of a network management system
                    that provides a list of information about
                    connections and disconnections and reasong
                    for these disconnections and excessive errors
                    on the network.

Baseband            An electrical signaling technique used to
                    transmit information.  Baseband signaling
                    uses unmodulated signals.  The carrier is
                    present only when data is being transmitted.
                    The entire frequency range of the channel is
                    used during this tranmission.

Broadband           An electrical signaling technique used to
                    transmit information.  Broadband signaling
                    involves modulation of the signal before
                    transmission.  Broadband networks typically
                    divide the total bandwidth of the
                    communication's channel into multiple
                    subchannels, so different types of
                    information can be transmitted simultaneously
                    using different frequencies.  Broadband
                    signaling is used when mixing multiple types
                    of information such as video, voice, and
                    data.

CCITT               The initials for the French International
                    Telegraph and Telephone Consultative
                    Committee.  This organization defines
                    standards or recommendations (e.g., X.25) for
                    international networking.

Coaxial Cable       A physical tranmission medium with two
                    conductors.  The center conductor carries the
                    information signals.  The outer conductor
                    (electrostatic shielding) acts as a ground.

Collision           (See CSMA/CD)
Detection

Communications      A hardware and software device that allows
Server              devices such as terminals, host computers, or
                    printers to access a network without having
                    to implement the communications protocol in
                    the device itself.  The communications server
                    communicates with the device using standard
                    protocols built into the device.

Connection          A communications path between two devices
                    that allows the exchange of information.
                    Other terms used to refer to a connection are
                    session or circuit.

CRC                 Abbreviation for Cyclical Redundancy Check.
                    This is a method of detecting errors in a
                    message by performing a mathematical
                    calculation on the bits in the message and
                    then sending the results of the calculation
                    along with the message.  The receiving
                    network station performs the same calculation
                    on the message data as it receives it and
                    then checks the results against those
                    transmitted at the end of the message.  If
                    the results do not match, the receiving end
                    asks the sending end to send the entire
                    message again.

CSMA/CD             Abbreviation for Carrier Sense Multiple
                    Access with Collision Detection.  It is an
                    access method that allows many nodes to share
                    a single channel of a communications medium.
                    If more than one signal is transmitted at the
                    same time, the signals collide and are
                    retransmitted at a later randomly calculated
                    time.

Datagram            A transmission method in which sections or a
                    message are transmitted in scattered order
                    and the correct order is reestablished by the
                    receiving workstation.

DECnet              Digital Equipment Corporation's proprietary
                    communications protocol.

Ethernet            A local area network that utilizes baseband
                    signaling at 10 Mbps.  The development of the
                    Ethernet specification was a joint effort by
                    Xerox, DEC and Intel and is the predominant
                    local area network standard.

FDDI                Abbreviation for Fiber Distributed Data
                    Interface.  FDDI is an emerging standard for
                    a 100 Mbps fiber-optic LAN.  It uses a
                    "counter-rotating" token ring topology.  It
                    is compatible with the standards for the
                    Physical layer or the ISO model.

Flow Control        The hardware or software mechanisms employed
                    in data communications to turn off
                    tranmission when the receiving workstation is
                    unable to store the data it is receiving.

Frame               A group of bits sent over a communications
                    channel, usually containing its own control
                    information, including address and error
                    detection.  The exact size and format of a
                    frame depends on the protocol used.

HDLC                Abbreviation for High-level Data Link
                    Control, which is the ISO procedure for data
                    link control.  HDLC uses a specific series of
                    bits rather than control characters for
                    transmitting and receiving data.

High-Level          A protocol that allows network users to carry
Protocol            out functions at a higher level than merely
                    transporting streams or blocks of data; for
                    example, reliably transmitting data,
                    formatting data, establishing a connection,
                    transferring a file.

IEEE 802.2          A Data Link layer standard used with the IEEE
                    802.3, 802.4, and 802.5 standards.

IEEE 802.3          A Physical layer standard specifying a LAN
                    with a CSMA/CD access method on a bus
                    topology.  Ethernet following the 802.3
                    standard.

IEEE 802.4          A Physical layer standard specifying a LAN
                    with a token passing access method on a bus
                    topology.  Used with Manufacturing Automation
                    Protocol (MAP) LANs.

IEEE 802.5          A Physical layer standard specifying a LAN
                    with a token passing access method on a ring
                    topology.  Used with IBM's token ring
                    hardware.

ISO                 Abbreviation for International Standards
                    Organization.  (See also ISO Model)

ISO Model           ISO has developed the Reference Model for
                    Open Systems Interconnection, which divides a
                    complex set of communications functions into
                    self-contained modules.

LAN                 Abbreviation for Local Area Network.  A LAN
                    is a communications network that provides
                    high-speed data transmission over a small
                    geographic area.

LLC                 Abbreviation for Logical Link Control.  Upper
                    sublayer of the Data Link layer of the ISO
                    model.

MAC                 Abbreviation for Media Access Control.  The
                    lower sublayer of the Data Link layer of the
                    ISO model.  The MAC layer supports medium-
                    dependent functions.

Network             The overseeing and maintaining of a network.
Management          The duties performed by a network management
                    system include installing and configuring the
                    network, maintaining an operation log,
                    monitoring network performance, and
                    statistics.

Network             Geography of a network or a set of networks.
Topology

Node                Point in a network where service is provided,
                    service is used, or communications channels
                    are interconnected.  Sometimes used
                    interchangeably with station.

Packet              A block of data handled by a network in a
                    well-defined format.

Packet              The internal operations of a communications
Switching           network that uses software to dynamically
                    route packets from a source to a destination.
                    Packet switching allows the sharing of a
                    single communications channel among several
                    connections.

PAD                 The abbreviation for Packet Assembler and
                    Disassembler.  A PAD is a device that allows
                    asynchronous terminals to have access to a
                    Public Data Network.

PDN                 A network that provides data transmission
                    services to the public.  Typically, a Public
                    Data Network uses packet switching
                    technology.

Protocol            A strictly defined procedure and message
                    format that allows two or more systems to
                    communicate over a physical transmission
                    medium.  Due to the complexity of
                    comunications between systems and the need
                    for different communications requirements,
                    protocols are divided into layers.  Each
                    layer of a protocol performs a specific
                    function, such as routing, end-to-end
                    reliability, and connection.

SNA                 Abbreviation for Systems Network
                    Architecture.  The network architecture
                    developed by IBM.

Subnet              A portion of a network that is partitioned by
                    a router.

T-1 Carrier         A digital transmission system developed by
                    AT&T that sends information at 1.544 megabits
                    per second.  T-1 links can transmit voice or
                    data.

TCP/IP              Abbreviation of Transmission Control Protocol/
                    Internet Protocol.  A set of de-facto
                    networking standards commonly used over
                    Ethernet or X.25 networks.  It was originally
                    developed by the U.S. Government and is now
                    supported by many equipment manufacturers.
                    It defines high-level protocols such as
                    Telnet (terminal connection), FTP (file
                    transfer), and SMTP (electronic mail).

Thin Ethernet       A ligher (0.2" diameter, black coating)
                    variation of Ethernet cable that saves cable
                    and installation costs, but is restricted in
                    effective distance.  This type of cabling is
                    specified under the IEEE 802.3 10Base2
                    standard.

Token Ring          A technology developed by IBM whereby a token
                    is used to direct traffic on the network.
                    There is only one token on the ring (the
                    network) at a time.  It is either free or
                    busy.  A node must wait for a free token to
                    tranmit data; it marks the token as busy,
                    then transmits a frame of data onto the
                    ring.  Data collisions cannot occur as only
                    one node can transmit at any one time.

Twisted Pair        A form of wiring commonly used for telephone
                    installations.  Standard networks such as
                    Ethernet can operate over such wiring.  This
                    method is economical, but poses distant
                    limitations unlike coaxial cable.

VAX                 A trademark name for a family of computers
                    manufactured by Digital Equipment
                    Corporation.

Virtual Circuit     A facility in a packet switching network in
                    which packets passing between a pair of
                    devices are kept in sequence.  This is a
                    "virtual circuit" because it appears there is
                    an actual point-to-point connection.

WAN                 Abbreviation for Wide Area Network.  A data
                    communications network designed to serve an
                    area of hundreds or thousands of miles.  A
                    WAN can be public or private.

X.25                A CCITT standard that defines the standard
                    communications protocol by which mainframes
                    access a public or private packet switching
                    network.  These networks are often referred
                    to as X.25 networks.

XNS                 Abbreviation for Xerox Network Systems.  A
                    protocol family specifically designed to run
                    on Ethernet.  It contains the Internetwork
                    Datagram Protocol (level 3) and the Sequenced
                    Packet Protocol (level 4).

