Calhoun: The NPS Institutional Archive

Theses and Dissertations Thesis Collection

1991-09

Local area network (LAN) compatibility issues.

Espiritu, Rita V.
Monterey, California. Naval Postgraduate School

http://hdl.handle.net/10945/28335
NAVAL POSTGRADUATE SCHOOL
Monterey, California

THESIS
LOCAL AREA NETWORK (LAN)
COMPATIBILITY ISSUES

by

RITA V. ESPIRITU

September, 1991

Thesis Advisor: Dr. Norman Schneidewind

Approved for public release; distribution is unlimited

T259717
 REPORT DOCUMENTATION PAGE
1a REPORT SECURITY CLASSIFICATION lb RESTRICTIVE MARKINGS
unclassified

2a SECURITY CLASSIFICATION AUTHORITY 3 DISTRIBUTION/AVAILABILITY OF REPORT

Approved lor public release; distribution is unlimited.
2b DECLASSIFICATION/DOWNGRADING SCHEDULE

4 PERFORMING ORGANIZATION REPORT NUMBER(S) 5 MONITORING ORGANIZATION REPORT NUMBER(S)

6a NAME OF PERFORMING ORGANIZATION 6b OFFICE SYMBOL 7a NAME OF MONITORING ORGANIZATION

Naval Postgraduate School (If applicable) Naval Postgraduate School
55

6c ADDRESS (City, State, and ZIP Code) 7b ADDRESS [City, State, and ZIP Code)
Monterey, CA 93943 5000 Monterey, CA 93943 5000

8a NAME OF FUNDING/SPONSORING 8b OFFICE SYMBOL 9 PROCUREMENT INSTRUMENT IDENTIFICATION NUMBER

ORGANIZATION (If applicable)

8c ADDRESS (City, State, and ZIP Code) 10 SOURCE OF FUNDING NUMBERS

Program t lement No Hroieci No Work Unit Accession

Number

1 1 TITLE (Include Security Classification)

LOCAL AREA NETWORK (LAN COMPATIBILITY ISSUES )

12 PERSONAL AUTHOR(S) Rita V. Espintu

13a TYPE OF REPORT 13b TIME COVERED 14 DATE OF REPORT (year, month, day) 15 PAGE COUNT
Master's Thesis From To September 1991 95
16 SUPPLEMENTARY NOTATION
The views expressed in this thesis are those of the author and do not reflect the official policy or position of the Department of Defense or the U.S.
Government.
17.COSATICODES 1 8 SUBJECT TERMS (continue on reverse if necessary and identify by block number)
FIELD GROUP SUBGROUP Local Area Networks(LAN), LAN hardware and software compatibility issues, LAN
Manager's Guide.

1 9 ABSTRACT (continue on reverse if necessary and identify by block number)

This thesis presents a research study of local area network compatibility issues. Todays high performance and successful LANs must use
hardware and software that is compatible with the network technology used. Compatibility, for the purposes of this thesis, means the ability to
programs within the network and
interface without special adapters or other devices and directly relates to the ease of the rnsfer of data or
between systems. An understanding of compatibility issues can help network users and managers diagnose and resolve connectivity problems
thus saving valuble time and money that can be used toward other productive endeavors within an organization. This thesis will discuss
hardware and software concepts, LAN architecture, and design issues as they relate to network compatibility. A LAN Manager's Guide covering
the above issues is included as an appendix. It is written specifically for the Administrative Sciences/Information Systems (AS/IS) computer
network laboratories as supplemental information for students on LANs.

20 DISTRIBUTION/AVAILABILITY OF ABSTRACT 21 ABSTRACT SECURITY CLASSIFICATION

|[j UNCLASSIFIED/UNLIMITED J SAMt AS REPORT J OTIC USERS Unclassified
22a NAME OF
RESPONSIBLE INDIVIDUAL 22b TELEPHONE (Include Area code) 22c OFFICE SYMBOL
Norman Schneidewind (408)646-2719 AS/Ss
DD FORM 1473. 84 MAR 83 APR edition may be used until exhausted SECURITY CLASSIFICATION OF THIS PAGE
All other editions are obsolete Unclassified
 Approved for public release; distribution is unlimited.

Local Area Network (LAN) Compatibility Issues

by

Rita V. Espiritu
Lieutenant Commander, United States Navy
M.S. Naval Postgraduate School, 1991

Submitted in partial fulfillment

of the requirements for the degree of

MASTER OF SCIENCE IN COMPUTER SYSTEMS MANAGEMENT

from the

NAVAL POSTGRADUATE SCHOOL

September 1991

n
 ABSTRACT

This thesis presents a research study of local area

network compatibility issues. Today's high performance and

successful LANs must use hardware and software that is

compatible with the network technology used. Compatibility,

for the purposes of this research paper, means the ability to

interface without special adapters or other devices and

directly relates to the ease of the transfer of data or

programs within the network and between systems . An

understanding of compatibility issues can help network users

and managers diagnose and resolve connectivity problems thus

saving valuable time and money that can be used toward other

productive endeavors within an organization. This thesis will

discuss hardware and software concepts, LAN architecture, and

design issues as they relate to network compatibility. A LAN

Manager' s Guide covering the above issues is included as an

appendix. It is written specifically for the Administrative

Sciences/Information Systems (AS/IS) computer network

laboratories as supplemental information for students on

LANs.

111
 TABLE OF CONTENTS

I. INTRODUCTION 1

A. BACKGROUND 1

B. PURPOSE 3

C. SCOPE AND METHODOLOGY 3

D. ORGANIZATION OF STUDY 4

II. NETWORK STANDARDS 5

A. DETERMINING COMPATIBILITY 5

B. NETWORK ARCHITECTURE 7

C. COMMUNICATIONS STANDARDS 8

1. ISO's Open Systems Interconnection (OSI)

Model 9

2. IEEE 802 Standards 12

D. MEDIA STANDARDS 12

E. DATA TRANSMISSION STANDARDS 13

F. LAN TECHNOLOGIES 14

1. EtherNet 15

2 IBM' s Token-Ring 15

3. ARCnet 16

4. Appletalk 17

G. LAN TOPOLOGY 18

H. COMMUNICATION PROTOCOLS 22

IV
III. NETWORK HARDWARE 25

A. NETWORK INTERFACE CARDS (NICs) 26

B. FILE SERVERS 31

C. MICROPROCESSORS /CENTRAL PROCESSING UNITS

(CPUs) 34

D. BUS CARDS 34

E. ROM BIOS 35

F. RANDOM ACCESS MEMORY (RAM) 36

G. WORKSTATIONS 36

H. EMULATION BOARDS 37

I. INTERCONNECTION HARDWARE 39

1. Bridges 40

2. Routers 4

3. Gateways 40

IV. NETWORK SOFTWARE 42

A. OPERATING SYSTEMS (MS/PC DOS) 42

B. NETWORK OPERATING SYSTEM (NOS) 45

C. LAN DRIVERS 48

D. NETWORK BASIC INPUT/OUTPUT SYSTEM (NetBIOS) . . 49

E. APPLICATIONS 51

F. INTERRUPTS 54

G. TSR (TERMINATE AND STAY RESIDENT) PROGRAMS . . 55

V. CONCLUSION 57

v
APPENDIX: AS/IS LAB MANAGER'S GUIDE 58

LIST OF REFERENCES 84

INITIAL DISTRIBUTION LIST 88

VI
 I . INTRODUCTION

A . BACKGROUND

The Local Area Network (LAN) is a technology that has

evolved to meet the needs of automating short distance

communication at high speeds of operation and relatively low

error rates. They are a particular form of data

communications that optimize hardware and software to support

the sharing of devices and information. Not only do LANs

support personal computer workstations but they are also being

used in conjunction with minicomputer and mainframe networks

as well as wide area networks (WANs) . Today, LANs have

emerged as a useful technology loaded with immediate

practicality and a promising future. The mere sharing of

expensive resources such as laser printers and high capacity,

high-speed mass storage devices among many users is a direct

economic benefit to management and equates to significant

savings. Businesses realize the importance of local area

networks in their daily activities and recognize that

networking is desirable for optimum company effectiveness.

Local area networks have expanded to global proportions

and the technology has exceeded the tasks it was originally

acquired to perform. In these days of fast-moving

technologies, compatibility issues have become important.
Users have realized that compatibility not only results in

connectivity, but also in interoperability. Starting a LAN

and interconnecting it with other LANs, wide-area networks

(WANs) , and metropolitan-area networks (MANs) to expand its

scope are complex tasks that are here to stay. Software and

hardware vendors are going to great lengths to ensure

compatibility with as many network products as possible.

Sometimes, however, claims and intentions of vendors do not

always translate into compatible machines and products, and

many network users and managers end up spending numerous hours

troubleshooting problems that affect the performance of the

network.

Compatibility means that a hardware unit or software can

perform its intended function when it is part of a larger
program system, like a PC being able to perform its functions
when connected to a LAN. If a PC is said to be IBM PC-

compatible, that implies that the PC can run the same software

and support the same add-on hardware as an IBM PC; but it may

also mean that the PC can run most IBM PC-compatible software.

So, there are also degrees of compatibility that must be

understood. As hardware and software vendor's products claim

compatibility with established products and standards, they

also attain popularity in the market in that they can claim

that their products can operate on older equipment as well as

newer models of the same equipment. This is called backward

compatibility and forward compatibility, respectively.

B PURPOSE

The purpose of this thesis is to research and discuss LAN

issues that LAN users and managers may encounter. It is the

intent of this work to be a ready reference and guide to

various network compatibility issues, both in hardware and

software. It will assist users and managers in the resolution

of difficult compatibility and connectivity issues without

delving too deeply into the intricate network technology. It

is also aimed at providing the reader with an understanding

of compatibility issues regarding LANs. A LAN manager's guide

is included as an appendix. It is written specifically as an

aid for LAN managers and as a supplement in LAN courses

C. SCOPE AND METHODOLOGY

The main thrust of this research is the investigation and

identification of LAN compatibility issues with regard to

available network hardware and software. An overview of

network concepts and terminology will be discussed as well as

its relationship to various network technologies. It is the

intent of this thesis to supplement a student's knowledge of

local area networks and generate interest and awareness of

network compatibility issues that may be encountered by a

network lab manager. The Lab Manager's Guide included in the

appendix will be treated in tutorial fashion to allow for

easy readability and understanding.

 This research is primarily a review of current books,

periodicals, and news articles on Local Area Networks, with a

principal focus on network hardware and software compatibility

issues

D. ORGANIZATION OF STUDY

In order to understand compatibility issues of local area

networks, some basic network issues will be discussed. A

chapter on network standards is considered appropriate. This

will cover determining compatibility; the different standards

governing communication, media, and data transmission; the

organizations involved in the development of standards such as

the ISO and the IEEE; a number of network technologies and the

topologies and communication protocols each technology uses.

Follow-on chapters will cover various network hardware and
software, describing each and discussing the relevant

compatibility issues that impact a LAN's performance. Lastly,

a LAN MANAGER'S GUIDE is included as an appendix.

 II. NETWORK STANDARDS

A. DETERMINING COMPATIBILITY
Compatibility is defined as the capability of a

peripheral, a program, or an adapter to function with or

substitute for a given make or model of computer. It is also

the capability of one computer to run the software of another

company's computer. It should be noted though that, to be

truly compatible, a program or device should operate on a

given system without modification; all features should operate

as intended, and a computer claiming to be compatible with

another should run all the other computer' s software without

modification
Standards have been set at some of the lower levels of

networking, but there are still implementation

incompatibilities; at higher layers, standards are still being

developed. Practically every layer has some sort of

incompatibility yet to be resolved. From problems with

physical layer bridging to incompatibilities at the

application level, myriad incompatibilities lie quietly

underneath the marketing hype about "interoperability" and

"true multi-vendor networking." (Schatt, 1991)

Many companies today develop software and hardware that is

compatible to certain systems, the most notable of which is

the clone to the IBM PC. It was a year after IBM introduced

its PC that the first IBM PC-compatible computer was released.

Numerous companies designed products that were to be 100

percent compatible with IBM software and accessory devices,

such as displays and printers. Although these were known

collectively as clones, conveying the connotation of "cheap

imitation, " many of the computers produced actually improved

on the original, and so the term "clone" is no longer fair.

(Doll, 1990)

The original IBM PC was designed to use off-the-shelf

components-- such as disk drives, microprocessors, and power

supplies--that non-IBM companies had developed already for

earlier personal computers. These same components could be

assembled by anyone with requisite technical know-how. In

addition, IBM purchased the PC's operating system, PC DOS,

from the Microsoft Corporation, which was free to sell

virtually the same system (MS- DOS) to clone manufacturers.

The only part of the computer that IBM actually copyrighted

was a small amount of internal programming code, which other

computer companies could emulate without actually copying.

(Green, 1988)

To counter the compatible market, IBM attempted to close

the architecture of its personal computers by the release of

the PS/2 series in 1986. The key feature in this series is

its Micro Channel bus architecture, which has certain

technical advantages over the method used to communicate data

within previous PC's. However, the Micro Channel bus created

a closed environment for PC add-on boards and accessories; the

older boards and accessories do not work on a Micro Channel

machine, and any company developing products for Micro Channel

machines needs a license from IBM. (Green, 1988)

There are MCA clones. Most compatible makers have not

emulated the Micro Channel standard, preferring instead to

stick with the tried- and-true PC architecture. To take full

advantage of the 32-bit bus structure of the Intel 80386 and

80486 microprocessors, these manufacturers (dubbed the "Gang

of Nine") have created a bus standard called Extended Industry

Standard Architecture (EISA) . Computers conforming to the

EISA standard can accept existing adapters while taking full

advantage of these powerful new microprocessors. (Cole, 1990)

B. NETWORK ARCHITECTURE
The architecture of a computer network is the design plan

of how the network's component parts work together. It is

what distinguishes one network from another. It describes how

the system or program is constructed, how its components fit

together, and the protocols and interfaces used for

communication and cooperation among modules or components of

the system. In short, network architecture defines the

functions and description of data formats and procedures used

for communication between nodes or workstations. It combines

the existing standards and protocols needed to provide the

services desired to create a functioning network. It is

therefore a standard; the network architecture defines the

rules of a network and how the components of a network can

interact. It can be broken into layers, with each layer

responsible for a certain task. When these tasks are

combined, they result in a service being performed by the

network. (Durr, 1989)

Two different computer networks can have many components

in common, but their architecture makes them different. For

example, two networks can use the same kinds of cables and

connectors, but still have very different architectures.

C. COMMUNICATIONS STANDARDS
The heart of networking is communication and effective

communication requires a set of standards. It is important

that products from different vendors are able to communicate

with each other. Standards provide manufacturers a measure to

follow in order to ensure compatibility between products in a

multivendor network. As a result of standards, many different

manufacturers can produce compatible and complementary

equipment thus reducing the possibility that one or a few

companies can monopolize a niche of the industry. Because

many manufacturers can produce compatible equipment where

standards exist, marketing is forced to emphasize quality and

value-added services rather than particular protocol

differences. This leads to competition and lower prices of

the product for consumers. (Madron, 1990) Manufacturers

often view their own implementation of communication protocols

as the best solution and, over time have attempted to

influence the industry.

Compliance with a standard is still a vexing issue that

causes much rancor between vendors and users. As standards

become more sophisticated, proving compliance becomes more

costly and complex. Any ambiguities in the specifications can

lead to conflicting interpretations in products.

It is important to note that standards are adopted as a

result of economic self interest. Users want to have more

independence from specific vendors and governments may

encourage standards to foster cheaper solutions and increase

efficiency, but vendors will adopt standards only when there

are sufficient economic incentives. (Miller, 1991)

Fortunately, organizations have been established to help

create standard industry protocols so that network devices can

communicate even if they are using products from different

vendors. One of these organizations is called the

International Standards Organization (ISO) . Another is called

the Institute of Electrical and Electronics Engineers (IEEE)

1. ISO's Open Systems Interconnection (OSI) Model

The OSI model was created by the ISO as a result of a

committee study on the compatibility of network equipment.

The model provides the base upon which manufacturers design

products and is a general reference framework for LAN

standards. It is a manufacturer- independent architecture for

data communication protocols. The goal is to have various

systems conform to the reference model and thus be able to

connect and share meaningful information.

The OSI reference model is essentially a seven-layered

hierarchical model which defines specialized functions. It

defines a universal architecture for interconnecting different

types of computer systems. The seven layers are Physical,

Data-Link, Network, Transport, Session, Presentation, and

Application. Each layer defines a layer or level of function.

The services that each layer supplies are defined by a range

of standards, which allows for different ways of performing

the same service. Compatibility of equipment can be defined
within a layer, or lower-level implementations can be hidden

to achieve compatibility at some higher level. Network

software is written so that each layer is aware of and

communicates only with the layer immediately above and below

it. (Durr, 1989)

Each function or layer performs tasks necessary for

different levels of communications between systems. The

layers allow for the division of network tasks and make it

possible for changes to be made to one or more layers without

affecting any other layers of the system. (Bullette, 1991)

The model provides a framework for the definition of services

and protocols which fit within the boundaries established. It

10
does not specify how functions must be performed. This level

of definition requires a defined implementation (protocol) for

each layer. Another description of the model is that it is a

frame of reference for open systems (refers to a network model

open to equipment from competing manufacturers) with

implementation details being left to other standards. It does

not imply any particular implementation or technology.

Adherence to the reference model does not mean that one

vendor's products can communicate with another vendor's

products

It is important to understand that an open system can

be OSI compatible without providing the initial source or

final destination of data. (Madron, 1990) In other words, an

open system need not contain the higher layers of

architecture. The IEEE 802 standard, for example, applies

only to the lowest two layers: Data Link and Physical. This

is often the source of significant confusion since, when we

talk of 802.3 (commonly called Ethernet) or 802.5 (Token

Ring) , the discussion sometimes proceeds to commentaries on

TCP/IP (Transport Control Protocol/Internet Protocol or XNS

(Xerox Network Protocol) or some other "protocol." TCP/IP and

XNS exist at layers beyond the Physical and Data Link

(although TCP/IP and XNS are not OSI standard protocols)

The OSI model is far from complete. The initial DSI

work placed major emphasis on the definition of the seven-

layer model and the protocols needed for application-to-

ll
application communication. Lack of management specifications

reduces the major power of OSI - unifying different systems

and networking environments. (Miller , 1991)

2 . IEEE 802 Standards

The IEEE has developed a set of standards describing

the cabling, physical topology and access schemes of network

products (i.e., IEEE 802. X standards). The standards

describe the protocols used in the lower two layers of the OSI

model. They do not go above those layers, thus using the

common name of an IEEE Standard (like Token-Ring) is an

incomplete response to the question "What network do you use?"

The correct reply should also specify the network interface,

including media and access protocol as well as the networking

software. (Jordan, 1990) Note that the 802.x standards do

not describe every popular network cabling and access protocol

scheme (ARCnet , for instance, is not a perfect fit)

D. MEDIA STANDARDS
The cables and other carriers of a network's transmission

signals are called media. Cabling is the physical element

that connects workstations to servers, LANs, WANs, and

dissimilar environments such as minicomputers and mainframes.

(Liebing, 1990) The lifeblood of any data network is its

cable. The most common kinds of cables are: coaxial cable

(coax) , which can carry high-frequency data for relatively

long distances but can be expensive; twisted pair copper

12
cables, which can be much less expensive than coax, but it

lags slightly in transmission speed and maximum distance it

can carry a signal; and fiber-optic cabling, which uses light

instead of electricity to carry a signal. In many ways,

fiber-optic cable is the ideal cable because it can carry

extraordinarily high-frequency data for miles, and it is very

hard to tap into without detection. It is however, expensive

and less flexible, and so harder to install.

E. DATA TRANSMISSION STANDARDS

These standards relate primarily to the way in which data

are transmitted on a medium. With baseband systems, signals

are transmitted at their original frequencies (i.e.,

unmodulated) . It transmits all signals through a single

channel, one at a time. In contrast, broadband, often used to

characterize communications based on CATV technology,

transmits multiple signals at once by subdividing its

transmission medium into channels. It is also characterized

by a large bandwidth (a network's range of transmission

frequencies) resulting in the capacity for very high data

rates. Transmission speeds for broadband range from one to

five mbps . The slower transmission speeds are acceptable for

some installations because of the significant increase in

cabling distance. Cables run can extend to as long as 50

kilometers. (Bullette, 1991) Broadband systems are modulated

and in their CATV context, are analog systems.

13
F. LAN TECHNOLOGIES

The technology of a computer network has two main aspects:

communication protocols and topology. This thesis will

discuss some popular network technologies for LANs: EtherNet,

Token-Ring, ARCnet, and Appletalk. The debate over the

performance merits of one networking technology over another

has been raging since the inception of LANs. (Liebing, 1990)

Traditionally, the debate revolves around raw transmission

speed. The EtherNet camp claims that its 10Mbps transmission

speed beats Token-Ring's 4Mbps and ARCnet 's 2.5Mbps hands

down. The IBM Token-Ring camp countered that by coming up

with a 16Mbps version. They also claimed that EtherNet'

CSMA/CD medium access protocol allows collisions on the wire

to degrade performance considerably as more workstations are

added to the network. The EtherNet proponents responded by

pointing out that Token-Ring' s token-passing protocol adds a

significant amount of overhead, and that when a small number

of workstations are passing a large amount of data across the

network, time spent passing a token is time that could be

better spent transferring actual data. Through it all, the

ARCnet community seems quite content with their slower,

simpler token-passing alternative. (Liebing, 1990)

Ultimately, the performance claims of any network

architecture must be weighed against a number of other

factors: ease of installation, cost per node, quality of

manufacturing, efficiency of associated LAN drivers,

14
availability of supporting hardware (repeaters, hubs, media

access units, etc.) , flexibility of the topology, cabling

types supported, adherence to standards, and compatibility

with other systems

1 . EtherNet

The original EtherNet networking scheme calls for a

thick coaxial cable, which has come to be known as thick

EtherNet cable. Following the classic bus topology, a single

thick EtherNet trunk cable runs from one end of the network to

the other. The stations are connected to the bus via

transceivers and transceiver cables.

Most PC-based EtherNet networks use the thinner RG-58

coaxial cable, which has been dubbed thin EtherNet or

"Cheapernet" cable. Stations are connected using a linear bus

topology in which cable is strung from one station to the

next. A T-connector is used to attach the cable to the
network interface board at the back of each network station.
EtherNet uses the CSMA/CD medium access method and provides a

transmission speed of 10 megabits per second (Mbps)

(Liebing, 1990)

2 . IBM' s Token-Ring

Its name applies a ring, however, the token ring is

physically a star, logically a ring. Media repeaters can be

added between two Multistation Access Units (MAU's) to

increase the transmission distance. Each node acts as a

15
repeater--receiving the serial bit from the nearest active

node, processing as necessary, and sending the bit stream on

down the cable to the next node in line. Only a few bit times

of delay are required in each workstation for these functions.

The serial transmission follows a complete ring or loop, with

the sending station eventually receiving its transmitted

information back after this information completes one round

trip around the ring. The token is merely a specific bit

sequence that circulates among the nodes, giving permission to

transmit. For this reason, these networks are often described

as being distributed polling environments. Each workstation

is thus polled to determine if it needs access to the network.

When the node is in possession of the token, it can transmit

a message that is in its output buffer. Otherwise, the node

is bit-repeat (and/or receive and process) mode. (Miller,

1989)

3 . ARCnet

Of all the major networks, ARCNET is the most flexible

in its architecture. Both star and bus topology networks, or

a combination that might be described as a distributed star

with branches, are possible. In addition, all three media

choices: coax (RG-62/U) , a single twisted pair (100 ohm

impedance, 2 twists/foot, 22, 24, and 26 gauge solid or 24 and

26 gauge stranded) and duplex fiber optic cable (50, 62.5,

100, or 200 micron core) are available and can be flexibly

16
interconnected. The only major constraint on the transmission

medium is that the signal propagation delay between any two

workstations must not exceed 31 microseconds. In addition,

the attenuation characteristics of the different types of

cable affect the number of workstations that can be attached

in a bus topology. (Miller, 1989)

4 . Appletalk

Appletalk is a baseband network with a bus topology.

It operates over shielded twisted-pair wire, unshielded
twisted-pair, or optical fiber--all at 230.8 kilobits per

second. Appletalk is a proprietary network standard that does

not conform to any IEEE specifications. But it uses a layered

architecture, and its protocols at every layer have been

published, so other vendors can develop products for it.

Appletalk networking uses an access scheme that is

similar to EtherNet's CSMA/CD. Under the CSMA/CA scheme, each

MAC on the network listens for a digital clock pulse that

indicates another station is transmitting. Appletalk is slow

because the zilog chip that runs it can be overwhelmed by

clock pulses that arrive at rates faster than 230.8 kilobits

per second. When data is sent any faster than that over the

network, MACs are prone to not detecting the clock pulse,

deciding there is no traffic, and barging into conversations.

(Needleham, 1990)

17
G. LAN TOPOLOGY

Topology refers to the physical layout of the cable; how

the network components are arranged in relation to each other.

Topology also refers to the access method of sending

information across the cable. It is the way in which network

interface boards arbitrate with each other for access to the

shared wire. (Liebing, 1990) Two methods are currently in

widespread use:
• Carrier Sense Multiple Access with Collision Detection
(CSMA/CD) - is most commonly used on an Ethernet linear
bus topology. A network board must wait until no one else
is talking on the network before it can talk. Then, as
the board sends the packet, it listens to what is being
said. If the board does not hear itself, another board is
talking at the same time. Both boards jam each other,
back off for a random amount of time, then try again.

• Token passing - Token passing is used on a ring, such as

IBM Token-Ring, and on a star bus, such as ARCnet It is .

more complicated than CSMA/CD. The network interface

boards pass a special "token" packet to each other.
Possession of the packet means that a board can send a
message if it wants. The trick to token-passing is
adapting the passing sequence when new boards are added to
the ring, and detecting when a board quits operating.

The three basic types of topology are the star, bus, and

ring topologies. On the OSI model, the cabling lies below the

first physical layer. One point must be made clear. The

actual installation of a LAN may not truly resemble the

topology used in its design. Although a network can be

physically laid out in a particular network topology, it does

not necessarily mean that is the route the information will

flow. Topology can be either physical or logical and both are

18
relatively independent. Both topologies can affect the

network's reliability, economy and resistance to interruption.

The nodes in a LAN handle messages in one of two logical

ways. Either they relay the messages from node to node in a

sequential logical topology or they send the messages out to

all stations simultaneously in a broadcast. EtherNet uses a

broadcast topology while Token-Ring uses a sequential

technique.

There are several ways of physically running the cables

connecting a group of computers. In the star or hub topology,

the network wires run between the network nodes and a central

wiring hub, usually located in the building's wiring closet.

The hub is the focal point of the star and all network traffic

must pass through the hub. Each workstation is like a spoke

emerging from the central hub, much like a wagon wheel.

The primary advantage of the star wiring topology is

operational survivability. The wiring hub isolates the runs

of network cabling. Even if a wire between a station and the

wiring hub breaks or develops a bad connection, the rest of

the network remains operational. It is also typically easier

to install than cable running from point to point since the

wire runs from the wall plates to a central point, like

telephone wiring. The overall installation is neater because

fewer wires run to each node. This topology makes it easier

to move PCs and change connections.

19
 A disadvantage of

using the topology would

be that it uses more wire

and there is the cost of

the wiring hub. Also, a

separate cable must be run

to every machine that

accesses the network, and Figure 1. Star Network Topology

if the hub goes down, so

does every station connected to it. (Derfler, 1991)

The second type of

topology is the bus, also

called daisy chain

topology. It is the

simplest of the wiring

schemes. (Derfler, 1991)

Figure 2 . Bus Network Topology The cable takes the

shortest path from one

network node to the next. The signal on the bus travels in

both directions from the workstations or nodes. Unlike the

ring topology, the data travels by the nodes and does not have

to go through each node. Instead, a coaxial T-connector

provides a tap into the cable at each network node. This

topology is typically associated with EtherNet

20
 Because of the electrical characteristics of this

topology, the following disadvantages exist: If the cable

breaks at any point, the entire network fails. The cable does

have many connection points, and a bad connection at any point

spells failure. The cable installations are also often messy

because two cables run from the CPU to the back panel of each

computer and then go off along the floor in separate

directions. Large networks can also saturate the bus when too

many machines try to communicate on the same bus "party line."

Additionally, security may be a problem since every station

can "hear" all network traffic. (Derfler, 1991)

The third basic

topology is the ring,

where each computer is

connected both to a

machine "upstream" and one

"downstream" from it, to

form a closed loop or a

ring. Each node on the Figure 3. Ring Network Topology

cabling acts as a

repeater, boosting the signal between workstations. Data

travels around the ring in only one direction and passes

through each node around the ring until it reaches either the

destination computer, which intercepts it, or it goes back to

the sender, where it stops.

21
 IBM' s Token-Ring is the most popular form of the ring

topology. While it is electrically a true ring, physically it

is normally wired as a star, with each computer trailing a

cable to a central box called a Media Access Unit (MAU) that

connects one cable to another. This kind of configuration is

called a star-wired ring.

Theoretically, a break in any cable on a ring network will

disable the entire network. But IBM's Token-Ring cabling

system has a built-in set of back-up cable, which reroutes

data back around the working part of the ring if a faulty

cable is removed from the network. Token-Ring cabling is

usually shielded twisted pair. (Madron, 1990)

H. COMMUNICATION PROTOCOLS
A protocol is nothing more than a formal agreement about
how computers should format and acknowledge information during

a communication session. (Derfler, 1991) When networking and

communications products from different companies follow the
same protocol, they have the ability to talk to each other, at

least theoretically. Protocols define the actual steps that

any device or program must take to communicate with another

device or program in the network. All systems must share the

same protocol type or go through a protocol conversion process

to be able to exchange information. The conversion device is

usually called a gateway and consists of both hardware and

software components. Gateways always add some overhead when

22
translating between protocols and therefore impact performance

somewhat. (Bullette, 1991)

Protocols specify everything from the transmission rate to

how data will be used by an application or end user. If any

part of the packet is not in a format recognizable by another

device, the two cannot communicate. Protocols govern format,

timing, sequencing, and error control. There can be sets of

protocols in some networks, with each protocol handling rules

for a subset of the entire task of communication. This set of

protocols is often called a "protocol suite" or a "protocol

stack." ( Rinzler, 1990)

In the context of the ISO seven-layer model, a protocol

refers to the rules associated with a specific layer or set of

layers. The protocol in any layer includes interface

standards for requesting service from the layer below to the

layer above. (Jordan, 1990)

Protocol rules are built into networking software, and

network devices observe protocols each time they perform a

task. In any network architecture, there must exist protocols

that define three generic classes of functions:

• Application Services - are the highest level network
functions. They enable an application program to
communicate with an equivalent program on another
computer.

• Transport Services - are lower-level network functions,

which manage addressing and other transmission control
tasks

23
 • Connection Services - are the lowest-level network
functions. These govern the actual, physical transmission
from one computer's memory onto the network, and then into
the destination computer.

A poorly implemented protocol can slow data transfer,

making very inefficient use of the network and degrade the

performance of the network. On the other hand, a well

implemented one can make communication between dissimilar

networks possible. The TCP/IP protocol provides for transfer

of data between computers with different architectures and

operating systems. As a network grows more and more

heterogeneous, matching protocols exactly is more difficult.

Vendors may develop protocols that vary slightly from the

published standard and from one another. This makes for

incompatibility in some instances. Other vendors may develop

proprietary protocols, independent of all standards, and yet

others combine protocols from various standards to take

advantage of the best features. (Jordan, 1990)

24
 III. NETWORK HARDWARE

Hardware requirements for a computer that will run network

operating system and application software can be complex.

Systems have to run the appropriate versions of software that

are compatible with the hardware (that is, network interface

cards (NICs) , disk controller cards, microprocessors, etc.)

The best computers must be able to mix hardware and software

from various vendors into a cohesive package. This is no easy

feat since most software and hardware products have bugs and

various incompatibilities built in, whether inadvertently or

deliberately

LAN hardware can be defined as the LAN's physical

components, along with the methods used to connect those

components. Cables, topologies, and network interface cards

(NICs) are the primary pieces of LAN hardware. CPUs, servers,

hard disks, and bus cards are other components. The selection

of each component depends on the physical layout and on the

performance requirements of users and applications. Because

not all types of cable can support long cable runs (i.e., if

long runs are necessary) , the cable choices are limited.

Transmission speeds, or raw bit rates, vary with NICs and

should be chosen to match the needs of the application. The

25
availability of diverse LAN hardware, interchangeable

standards, and the LAN's distributed architecture, contribute

to the flexibility of the LAN.

A. NETWORK INTERFACE CARDS (NICs)

NlCs refer to a local area network board, also called a

LAN adapter card/board. It provides intelligence and

connectivity to the network cabling system, using any of a

number of different wiring schemes currently available. The

NIC establishes most of the LAN's hardware characteristics.

(Durr, 1989) These characteristics include the cable type,

topology, access scheme, and data transmission (bit) rate.

Every computer attached to a LAN uses one sort of network

interface card or another. In most cases, the card fits

directly into the expansion bus of the computer. In some

cases, the card will be part of a separate unit to which the

computer attaches through a serial or parallel connection.

Dedicated workstations and servers already contain adapter

cards which are built into their motherboards.

Although the NIC provides the physical connection to the

network, the transfer of data from the PC to the network is

not that simple. To get one byte of data from here to there,

several things must happen. Buffers must be checked.

26
Requests must be acknowledged. Sessions must be established.

Tokens must be sent. Collisions must be detected, etc.

Luckily, the work of the NIC can be broken down into eight

tasks: host-card communications; buffering; packet formation;

parallel-serial conversion; encoding/decoding; cable access;
handshaking; and transmission/reception. These are the steps

taken to get data from the memory of one computer to the

memory of another computer. (LAN Magazine, Dec 88)

When choosing a NIC, such issues as performance, access

method, and topology must be considered. Reliability is also

an important issue. It does not matter how fast the NIC is if

it does not work, causes errors, loses packets, drops the line

etc. Isolating these type of network problems can be very

frustrating.

Most NICs are available in either eight-bit, 16-bit or 32-

bit varieties. An eight-bit card moves data from the

computer's I/O bus to the cable eight-bits (one byte) at a

time; a 16-bit card moves data 16 bits (two bytes) at a time.

Because of the higher performance of 16-bit cards, they are

most often used in the file server, whereas the slower eight-

bit cards are used in workstations. (Madron, 1990)

For compatibility purposes, the 16-bit card has two edge

connectors (one long and one short) so they will only fit in

a 16-bit slot in the expansion bay of the computer. One must

27
ensure that the file server has an adequate number of 16-bit

slots available to accommodate the boards that must be

installed. Also, many 16-bit cards can use hardware

interrupts nine or above. This capability is important in

avoiding hardware conflicts when the card must coexist in the

same machine with other expansion cards that use the lower

hardware interrupts. (Liebing, 1990)

The most common conflict between NICs and your PC occurs

when the NIC tries to use the same hardware interrupt (IRQ) as

another card or device in the PC, such as a mouse, internal

modem or serial port. Reading the manuals of the different

devices and noting which interrupts they use can prevent this

conflict during installation of the NIC. Also, jumpers on the

NIC will indicate which IRQs they use. Usually, if the system

uses COM1 for a modem, it is using IRQ4 . COM2 occupies IRQ3

Some NICs default to IRQ3, and if this is a conflict, a

command could be issued telling the card which IRQ to use.

The subject of interrupts is necessary to understand
compatibility issues and will be discussed in detail in a

separate section.

Another conflict noted is when communication errors

(duplicate requests, packet collisions) occur and it is

certain that cabling is not the culprit. There is a good

possibility that the system is being slowed down by the NIC.

28
Also, if the CPU is running 50 percent, there might be a need

for a new server with a 38 6 chip running a minimum of 25 or 33

Mhz. (Liebing, 1990)

The network operating system (NOS) communicates with the

NIC through a specialized piece of software called a

communication driver or LAN driver. This driver must be

linked to the operating system during installation. Due to

different protocols and different card designs, each NIC model

may have a unique driver. Unfortunately, there is very little

standardization in this area. For example, each EtherNet and

Token-Ring model has a different driver. ARCnet is the

exception: the ARCnet card manufacturers adhere to self-

imposed design standards that make these cards and their

drivers practically interchangeable. (Miller, 1989)

Different types of NICs are available depending on the

type of network. The most widely used NICs are EtherNet,

ARCnet, and Token-Ring.

General incompatibility problems with network boards and

disk controller boards include: (Foster, 1991)

Bus Mastering - Bus mastering is a fairly recent

concept supported by IBM' s Micro Channel Architecture (MCA)

and the new EISA technology (Enhanced Industry Standard

Architecture) , as opposed to ISA (Industry Standard
Architecture, which is known as the AT-type compatible bus)

29
The MCA or EISA buses are two or four times wider than that of

the ISA, and increase throughput significantly. Bus mastering

allows NICs to bypass the CPU and take over the bus. This

enables the system throughput to be a lot faster, since the

CPU no longer has to play the middleman, relaying commands

from the NIC to the bus and back. They both offer a burst

mode that lets data transfer from bus cards to memory (and

vice versa) and take place at the speed of the processor. Bus

mastering is not supported by the AT-type computers. (PC

Magazine, 29 May 90)

The difficulty comes in implementing the technology

uniformly; bus mastering intelligence has to be on both the

board and the computer. Since the board and the computer are

usually made by different manufacturers, and since bus

mastering is not a universally demanded feature yet, bugs have

been popping out. (Foster, 1991)

Configuration Option - No two boards in the same

computer can have the same memory addresses or IRQ's. Boards

have a configuration option which provide the information on

these unique settings. Board manufacturers provide tables to

help ensure the uniqueness of the board' s configuration
option; the default configuration option usually works if
there is only one board in the system, requiring no more
effort than double checking the manufacturer's settings.

30
 Occasionally, computers and network interface boards

do not work together. Sometimes this is because the system is

technologically more advanced than the board. A number of 486

computers are simply too fast for most of the boards out

there. One solution to fast computers and slow boards is to

slow down the computer by disabling the internal cache. This

defeats the purpose in buying the machine, but at least the

board will work. The better solution is to speed up the

boards, and this is currently being implemented. (Foster,

1991)

Remote Booting - In diskless workstations, LAN

boards cannot complete the physical connection to the file

server at booting time without an additional chip that

supports the remote boot for a workstation. Most LAN boards

are sold without this chip, although one can be purchased.

Some boards, usually the fast ones, are designed exclusively

for use in file servers; they do not even come with the chip.

Occasionally, if the chip is on the board, the workstation

does not support it.

B. FILE SERVERS

A file server is usually the fastest, most powerful

computer in the network. It has more memory on the

motherboard, more expansion slots so that more options can be

31
supported, and more and larger hard disk drives than any

workstation would need. Usually, it has two floppy disk

drives and a switch on the back to change the wait state of

the system for fast or slow software.

The file server functions as the central repository of

data and/or application programs for the network. It also

performs network traffic management functions and provides

security for the data. It performs only information

retrieval. It does not perform computations or specific

record searches within a database. (Nunemacher, 1990)

The basic issues to consider in determining compatibility

of file servers include the manufacturer, the CPU, the bus

type, the ROM BIOS, and the amount of RAM supported.

(Liebing, 1990)

There are compatibility issues between the file server and

the network interface board. In a file server, there will

always be at least one other board to contend with in

addition to the network interface boards, and that is the disk

controller board which manages the interface between the hard

disk drive and the computer. The file server disk controller

board is usually not interchangeable with a disk controller

board in a workstation. For one thing, it is going to be more

powerful. (Foster, 1991) There could be conflicts if there

is more than one disk controller board in the system. In its

32
simplest form, the hard disks would be on one channel; a disk

subsystem connected to another disk controller board in the

system would be on a second channel, and so on.

Most servers today are designed and used primarily for

file and print service. (Needleman, 1990) Its capacity and

speed are the two biggest factors in the growth and

performance of many LANs. One of the ways to get more out of

a server is to use a faster CPU. A 286 machine may be fine

for 20-30 users, but bigger networks require more horsepower.

The 486 is the server processor of choice. However, a

faster processor alone, is not enough. The rest of the server

system, particularly the disk drives, must be fast and large

enough to keep up with the pace of the CPU. Many of today's

servers can hold over a gigabyte of disk storage, and average

access times are falling below 10 milliseconds.

The fastest processor and disk drives will not do much

good if the data from the server trickles out to the LAN.
Today's common, dumb, eight-bit network adapters are

inadequate. Servers should have a 32-bit bus that can support

intelligent, 32-bit disk drives controllers and LAN adapters.

Bus Master adapters (boards capable of taking over the bus)

offer the best performance. (Byte Magazine, June 1990)

33
C. MICROPROCESSORS/CENTRAL PROCESSING UNITS (CPUs)

At the heart of every server and workstation is the CPU.

Every bit of data which enters or leaves the computer passes

through the CPU to be processed or redirected. IBM and

compatible PCs use the Intel 808x or 80x8x based processor

(8088, 80286, 80386, 80486) . The performance of each of these

processors varies from between .5 million instructions per

second (mips) to over 4 mips.

The CPU's clock speed specifies the rate at which a

processor can carry out instructions. Clock speeds vary from

4.77 mhz (million cycles per second) to 33 mhz or higher.

(Bullette, 1991)

A faster CPU and clock speed will yield better performance

from file servers. A 386 will be faster than a 286. Also,

consider the number of wait states used to match the

performance of the CPU with the rest of the file server

hardware. The lower the number of wait states, the faster

the server. (Liebing, 1990)

D. BUS CARDS

The system bus is the circuitry which connects all of the

various subsystems together. IBM compatible PCs use either of

the following types: (Jordan, 1990)

• AT Style - often referred to as the Industry Standard

Architecture (ISA) is found in most IBM PC AT and

34
 compatible computers. This bus accepts standard 8- or 16-
bit network and disk interface cards. The standard speed
is 8Mhz

• Micro Channel Architecture (MCA) - bus design found in

IBM's PS/2 model 50 or above. These PS/2 computers will
accept only boards designed for MCA. MCA cards will not
work in standard AT bus machines. MCA offers higher data
transfer speeds than the standard AT bus.
• Extended Industry Standard Architecture (EISA) - designed
as an alternative to IBM' s proprietary Micro Channel
Architecture. Most EISA machines are very high
performance computers specially designed to be used as a
network file server. They offer a higher transfer speed
than the AT bus, and they have the added advantage of
being able to accommodate cards designed for the AT bus as
well as EISA-designed cards.

E. ROM BIOS
On a PC, the ROM BIOS (Read Only Memory Basic Input/Output

System) contains basic instructions and procedures used by the

CPU. The BIOS is the interface between the computer's

hardware circuitry and the software (both the operating system

and applications) running on the machine. Among other things,

the ROM BIOS controls the computer's interaction with the
monitor and the keyboard.

Compatibility problems are encountered when file servers

with older ROM BIOS firmware experience difficulty (such as

the inability to boot, random crashing, and APPEND errors)

when running nondedicated versions of network operating

systems (NETWARE) . The ROM BIOS must be 100 % IBM compatible.

Also, the hard disk must be included in the ROM BIOS setup

35
tables or there will be trouble getting the file server to

recognize the disk when the setup routine is run.

F. RANDOM ACCESS MEMORY (RAM)

A practical minimum for RAM is at least 2MB. The maximum

amount of memory a file server can have is 16MB. Adequate

memory becomes especially critical when the server must

service a large number of users, or when it has more than 7 9MB

of hard disk storage space to keep track of. Four to eight

Megabytes is usually sufficient. (Liebing, 1990)

If extended or expanded memory is desired for a PC, a

memory-add-on-board can be purchased. To avoid potential RAM

problems, all of the RAM chips installed on a given memory

board must be the same type, or at least have the same speed

rating. (Liebing, 1990)

G. WORKSTATIONS

The most common workstation problems occur when other

devices are added which cause a conflict with the network

adapter board. Every adapter requires a unique interrupt

(IRQ) , direct memory access address (DMA) , and input /output

(I/O) base address setting. Since many devices use the same

settings as the adapter, it is necessary to make changes so

that each device contains nonconflicting settings.

36
 Users who install applications in their private

directories on the network may be unaware that some

installation processes overwrite network drivers on the

startup diskette or hard disk. These kinds of problems can be

alleviated by placing write protect tabs on the diskettes or

using the DOS ATTrib command to protect specific files from

modification or deletion. (Bullette, 1991)

H. EMULATION BOARDS
An emulation board is a circuit board that when combined

with emulation software supports the physical connection

between a PC and a Mainframe system. It enables the PC,

acting as an intelligent terminal to manipulate mainframe

data. One of the most widely used systems of intelligent

terminals is IBM's 3270 series. The 3270 terminals are used

in on-line (interactive) sessions with IBM host computer.

Through emulation hardware and software, the PC

workstation can perform the same functions as the 3270

terminal. In addition, the PC can store data from the host

computer, modify or reformat display data, and run local

applications programs that analyze the data.

All emulators do not work with all IBM compatible PCs.

The emulator board should fit any slot and be used like any

other circuit board. Other parts of the existing system, such

37
as cluster controllers, may be "IBM-compatible" but not

compatible enough to support a particular emulator without

some modification. Compatibility is very important. Each

key of the emulator-equipped PC should be evaluated.

Sometimes, there is considerable variation in what a key

actually does, what the documentation says the key does, and
what the comparable key on a 3270 does. An advantage is

emulation software that maps the keyboard to a preferred one.

When evaluating the software, make certain that the software

supports not only the two-way transfer but also the mainframe

environment - TSO, VM/CMS, or both.

The emulator must be totally compatible with the PC and

network, which means it should be possible to disconnect the

coax and connector from the 3270 terminal and plug the cable

directly into the emulator board on the PC. Then, the system

can be started with the emulator software and be connected to

the mainframe

Another concern is that the emulator settings must not

conflict with any other devices within the PC workstation, or

else the system will not function properly. As long as the

interrupts and I/O channels are not duplicated, other

communications software can be used, including networking

software.

38
 The PC to Mainframe link is fundamental to the integration

of PCs into organizations. It enables distributed resources

to be accessed from any point on the network, regardless of

vendor environment or machine type (mainframe, minicomputer,

or PC) . (Durr, 1989)

I. INTERCONNECTION HARDWARE

While bridges, routers, backbones and the like are able to

provide the capability to link multiple LANs, they cannot link

LANs that use incompatible technologies. If one department's

LAN uses EtherNet, while another uses Token Ring, there will

be a need for bridges and routers or gateways to link such

dissimilar LANs. EtherNet and Token Ring both use the same

logical link control (LLC) protocols (IEEE 802.2), so bridges

and routers between the two must understand the LLC protocols.

Sharing the same LLC protocols often is not enough,

because the LANs to be linked may be running different higher-

level network software or protocol stacks, (Ex. LAN Manager,

a network operating system, on Token-Ring works with a NetBeui

protocol stack, while LAN Manager on EtherNet may use a

version of IPX) . Even though they use the same network

operating system, there will be a need for more than a bridge

or router that only understands LLC protocols to get these

LANs to communicate. (Woram, 1990)

39
 When two networks are being connected, the connections are

made at the first identical OSI layer. For example, if layers

1 and 2 of the OSI model on the networks are different, but

layers 3 and up are the same, the connection is made at layer

3 by having identical layers talk to each other within the

connecting device.

Devices called repeaters achieve the lowest level of

interconnection. A repeater extends the distance of a LAN.

Other hardware used for interconnection are: (Jordan, 1990)

1 . Bridges
A bridge connects two networks that use the same

communication technology (e.g. Ethernet). It is the next

level of interconnection after a repeater.

2 . Routers
A router is a device that interconnects the bottom

three layers of two networks.

3 Gateways

Gateways connect totally dissimilar networks. They

may perform protocol conversion in all seven layers of the

OSI model. A gateway not only performs the router's functions

(determines where the packet is going) , but converts the

message from one packet format to another or from one data

code system to another. A common use for a gateway is to

connect a LAN and a SNA mainframe computer, changing protocols

40
and retransmitting packets between two entirely different

systems. When communication occurs through a gateway to a

dissimilar system, some functionality usually is lost. File

transfer between systems or terminal emulation at the PC

currently are the extent of connection functionality.

41
 IV. NETWORK SOFTWARE

This chapter will look at the operating system software

required for each workstation to manage its hardware resources

as well as some other significant network software. Operating

systems have programming interfaces, software programs that
make it easier for programmers to write application programs

that require operating system resources. Some of these

interfaces will be discussed as well as terminology associated

with network software, (NetBIOS, Interrupts, TSRs etc) as to

their importance to networks and issues involving

compatibility

A. OPERATING SYSTEMS (MS /PC DOS)

An operating system (OS) is a group of programs that

manage all computer system resources and operations. Until

the operating system is loaded and running, the computer can

do nothing. The operating system's responsibilities include

managing both internal memory (RAM) and the auxiliary storage
resources such as a floppy disk drive or hard disk drive. It

is responsible for scheduling the performance of tasks,

handling input/output requests, and monitoring system security

42
to ensure that users are authorized to view certain files and

use specific programs. An operating system also includes

utility programs that help the user perform key tasks such as

preparing a new disk for use with a specific computer

(formatting) and placing some key system files on a disk that

enable the user to copy files and delete material. (Schatt,

1991)

Operating system programs facilitate communication among

the computer user, the computer's hardware and software, and

any peripherals. In 1981, IBM and Microsoft Corporation

published MS-DOS, an operating system for the new IBM PC and

all other compatible PCs that used the Intel 8088 or 8086

microprocessor. IBM also sold a version of the operating

system labeled PC-DOS. Both operating systems were developed

for one user performing one task at a time (single user/single

tasking) . The major parts of MS-DOS can be described as the

supervisor, the input/output manager, the file manager, and

the command processor. The supervisor is responsible for

scheduling and coordinating the activities of all programs

running under the operating system. The input /output manager

is responsible for all information transferred to and from all

peripheral devices, including printers, bar code readers, disk

drives, and so on. The file manager's responsibility includes

all areas of file management, such as saving, loading,

43
deleting, copying, naming, and renaming files. The command

processor checks commands from the keyboard before forwarding

requests to the supervisor. It receives messages back from

other parts of the operating system and translates them into

language that the user can understand.

The OS in conjunction with the ROM BIOS, controls the

basic input /output (I/O) functions that take place inside the
computer. Input from the keyboard and other sources is

handled by the OS, as is the output to the monitor, a printer,

or other devices

The disk and file management routines within the OS handle

all disk I/O. These include creating, opening, closing,

copying, deleting, and renaming files; reading from a file,

writing to a file, and searching for files; and creating,

deleting, and changing directories. The OS is also

responsible for coordinating access to the CPU's computing

resources and allocating memory for various applications. It

provides a uniform method for initiating and terminating all

other programs that run on the computer. It also has the

capability to interpret and process commands typed at the file

server keyboard.

Although DOS is serviceable for most network users, it has

some major limitations. A DOS workstation can access only 640
kilobytes of RAM. This falls short for many programs today.

44
Workarounds such as expanded and extended memory serve some

applications for program or data handling. However, the

workaround software that allows DOS to access larger RAM is

neither efficient, nor is it compatible with all software.

Another limitation of DOS is the fact that it does only one

task at a time. To solve this problem, IBM and Microsoft

developed 0S2
0S2 does not work on old PCs (with 8088 or 8086 CPU's),

but instead is intended for the newer 80286, 80386, and 80486

processors

B. NETWORK OPERATING SYSTEM (NOS)

The NOS coordinates the LAN's hardware activity, just as

DOS coordinates hardware activity in an individual PC. The

NOS is often loaded after the computer's operating system.

System hardware provides the data paths and platform on the

network, but the NOS controls everything else. The NOS

intercepts the data that is entered at the PC and passes it

along the network channel, if appropriate, or passes the data

back to the local PC for processing. (PC Today Magazine, Aug

90) Functionality, ease of use, performance, management, data

safety, and security all depend on the NOS.

Today, interoperability between operating systems are

available and improving. In 1989, network operating system

45
companies fueled the growth of networks by announcing and

delivering products conforming to open standards instead of

proprietary protocols. AT&T, Digital, and 3COM led the

industry in providing interoperable products for open

standards. Instead of trying to lock-in and control each

account with unique communication standards, they are

delivering software that works according to nationally

accepted standards
Network operating systems available today are NOVELLE
NetWare, IBM PC LAN Program, IBM LAN Server, 3COM 3+0pen,

Banyan Vines, and Apple AppleShare. (Durr, 198 9)

A NOS is not one program but rather a series of programs.

Some of these programs run in the PCs acting as servers of

various types, and others run in PCs acting as client

workstations. The networking software in servers provides and

controls multiple simultaneous access to disk drives,

printers, and other devices such as modems and facsimile

boards. The networking software in client stations interrupts

and redirects the requests for service that application

programs generate and sends each of them to the appropriate

server for action. (Derfler, 1991)

A NOS is the software necessary to integrate the many

components of a network into a single system to which an end-

user may have access. It manages the services necessary to

46
ensure that the end-user has error-free access to network

resources. Its basic function is to manage network addressing

- provide the end-user the capability to establish a

communications link from one node to another. Additionally,

a NOS will normally provide a user interface that is supposed

to reduce the trials and tribulations of using the network.

Within the context of the NOS, applications such as an

electronic mail system can be written which allow virtual

circuits among network entities to be established without

direct human intervention. (Madron, 1990)

The choice of NOS, perhaps more than any other LAN

component, determines the success of the LAN. Criteria used

to evaluate network operating systems include the following:

• Performance - Response time or throughput in accessing

network resources.
• Functionality - This is related to each site's
requirements
• Compatible Applications - The NOS must be compatible to
the software used.
• Extras - The enhancements that the NOS can support.

• Cost - Costs are measured against specific tasks and short

term goals or against corporate long-term goals.
At the heart of the NOS is a concept called redirection -

taking something headed in one direction and making it go in

a different direction. The redirection software in each

client computer makes the resources available on the network

47
look like local DOS to the program and people using them.

Commands sent from the keyboard and from programs to drives

with names like D, E, and F are redirected over the network to

the appropriate file servers. Similarly, programs sending

output to a network printer address a local LPT port just as

they normally do. The print jobs are redirected to the shared

printer and queued on the PC acting as a print server until

the printer is ready to take the job. (Derfler, 1991)

C. LAN DRIVERS

LAN drivers are often called communication drivers because

they make communication over the physical cable possible.

They provide the link between the operating system software

and the network interface board, which is directly attached to

the network cabling.

Information or data packets go through the network and

they have a specific structure. A typical packet consists of

a header and error checking codes as well as the actual data.

The exact size and position of a packet's fields varies

depending on the type of network (ARCnet, Token-Ring,

EtherNet) . Because of these differences in packet structure,

network interface boards can communicate only with other

boards of the same type over a certain type of cable. If the

board is an EtherNet board, the driver performs actions

48
necessary to create an Ethernet packet. The LAN driver code

is also responsible for receiving packets and checking them

for internal consistency.

If the driver code is slow or inefficient, the performance

of the entire system suffers. Regardless of how fast and

technologically advanced a network interface board is, it

cannot live up to its full performance potential without a LAN

driver specifically written to take advantage of the new

capabilities of the hardware. For example, IBM's new 16Mbps

Token-Ring boards support packet sizes up to 18KB. However,

some drivers written for these boards transmit data in 1KB

packets, thereby limiting the performance of the new boards.

(Liebing, 1990)

D. NETWORK BASIC INPUT/OUTPUT SYSTEM (NetBIOS)

NetBIOS is an interface, not a protocol per se. It

contains some critical software that functions as an interface

between a program and network software on an IBM PC or

compatible. It accepts commands from the program via software

interrupts. Before using NetBIOS or making a NetBIOS call,

the program creates and fills out a sort of electronic form

called a network control block (NCB) . The NCB tells the

network software which of the 19 NetBIOS functions is to be

performed. (Needleman, 1990)

49
 It defines a series of commands, but not how these

commands are to be communicated between two stations on a LAN.

IBM did not completely explain what the NetBIOS commands do

under all conditions. This may have led to incompatible

NetBIOS implementations; programs that worked with one

implementation would not work with another. Another common

compatibility problem is user' s misconception that NetBIOS is

a protocol, rather than an interface. There is no guarantee

that two network adapters driven by two different vendor's

NetBIOS implementations will be able to exchange data

reliably; in fact, the odds are that they will not. Vendors
will implement NetBIOS on various protocol "stacks" (such as

Novell's IPX, Xerox's XNS, TCP/IP etc). Programs written for

a NetBIOS interface will generally port to various networks

with no major problem. However, mixing protocol stacks on the

same LAN may prevent NetBIOS programs from communicating with

each other. To make two adapters communicate via NetBIOS,

they have to be driven by compatible software, preferably the

same release from the same vendor. (Needleman, 1990)

The NetBIOS interface provides key information on

connecting together workstations, sending, and receiving

information, and terminating communications. It was designed

to be a low-level interface that works adequately,

particularly if all workstations are using the same set of

50
network protocols for routing information. Because it is a

low-level data interface, it needs to be linked to a common

protocol such as OSI or TCP/IP to ensure that network adapter

cards are compatible. You can also use products that bridge

together NetBIOS interfaces linked to different protocols.

The NetBIOS was a revolutionary service that supported a

connection between nodes on the same network. Today, its

features seem limited because the system has no implicit

support for internetworking. It addresses only a small range

of functionality required by sophisticated applications.

(Durr, 1989)

NetBIOS is also a protocol independent session layer

interface to which programmers can write for peer-to-peer

communications. It provides an addressing scheme that allows

nodes to be identified by a name rather than an esoteric

number. These names can then be called to create a session

between the nodes. Once the session is established,

datagrams can be sent through the connection.

The service is very useful to communicate with another

node without going through the network servers. In its native

format it does not internetwork well.

E . APPLICATIONS
Applications are the real purpose for which networks

exist. Without powerful, network-aware, multiuser

51
application programs, LANs are just an expensive way for

isolated users to share disk storage and printers. Putting

applications on the network should be the overriding basis for

all network setup and maintenance decisions. Ultimately,

everything relates to what application is running and how it

is set up.

A surprising number of LAN problems are related to the

improper or careless installation of application software.

Most network applications are fairly easy to install on the

file server; others are not so easy. Then, there are always

those application programs for which a network version is not

yet available. (Liebing, 1990) To determine the degree of

network compatibility, the following major characteristics of

the application must be considered: (Liebing, 1990)

• Single-user versus Multiuser Applications - single-user

applications are usually modified to run in a multiuser
environment and they do not take advantage of the
distributed processing capabilities inherent in multiuser
applications
• Server-Based Applications - In this type of application,
the workstations become "clients" of a "server" that
performs services other than the usual file services of
the traditional file server. The services provided by a
server-based application can either be something the
workstations cannot do for themselves (usually the
functions of a print server, CD- ROM server, or
communications gateway involving specialized hardware not

52
 locally available to the client) or something the
workstations could do for themselves but not as
efficiently as the server.
• Multiple Defaults and Setup Files - A common difficulty
with applications that are not very "network-aware" lies
in the way they handle user preferences for defaults,
configurations, and setup options. Most true multiuser
applications provide a way for users to save their
preferences in their personal network directories so that
they do not overwrite other user's preferences.
• Network Operating System Features - The easiest
applications to install are those that have a version
designed specifically for a particular Network Operating
System.

• License Agreements - Terms of the application' s software

license agreement. Multiuser software is usually sold
with one of three licensing schemes: a site license, a
per-server license, or a number-of-users license. A site
license allows an unlimited number of users to run the
application simultaneously as long as they are at the
licensed site. With a per-server license, one copy of the
software is loaded onto the file server. It cannot be
loaded on any other server. Under a number-of-users
license agreement, a user can purchase a version of the
software tailored for a specific number of users.
• Fitting the Application into the Directory Structure -
Some application programs require that all program files
go into a single directory; with others, only certain
files go into a main directory, and the rest go into
nested subdirectories. This would make it easier to
isolate files for back-up, and make it easier to upgrade
or delete an application from the network.

• Share Critical Files as "Read-Only" - This prevents users

from writing over program files. Making executable files
"Read-Only" also helps prevent viruses from attaching
themselves to the files. Share .EXE and .COM files should
also be "Read-Only."

• Conserving Drive Mappings - certain applications require

more than one search drive mapping to function properly.
Often, one search drive must be mapped to the directory
containing the actual program files, and another must be
mapped to a directory containing specific configuration or

53
 setup files. Other programs may even require four or five
search drive mappings. Also, every drive mapping users
establish takes up 16 bytes of file server memory.

F . INTERRUPTS

An interrupt is a signal from a device or program that

halts processing momentarily so that input/output or other

operations can take place. When a program is interrupted or

temporarily suspended, control is transferred to the

supervisor. Interrupts use IRQ (interrupt requests) lines,

registers, stacks, and parameter blocks to accomplish

operations and they have priority levels, and higher-priority

interrupts take precedence in processing. (Derfler, 1991)

When the operation is finished, processing resumes. There are

two types of interrupts: hardware interrupts and software

interrupts. In a hardware interrupt, the signal is generated

by a device (e.g., disk unit) . In a software interrupt, the

program generates a signal that suspends processing so that a

specific operation can take place. (Pfaf fenberger, 1990)

LAN cards/boards in a system must each be set to an IRQ

that is unique to them. This is called the interrupt setting

and it is the most challenging setting on the card and the

cause of many network compatibility problems. I/O address

conflicts and hardware interrupt conflicts are the most likely

causes of a malfunctioning computer on a network. One of the

54
problems with the PC and XT is the shortage of interrupt

lines. They have only eight interrupt lines apiece, handled

by one interrupt controller. The AT has two interrupt

controllers working together for a total of 15 interrupts, but

most LAN adapter cards have a limited selection. Examples of

some interrupt functions are:

• Interrupt 3 is COM 2, the second serial port.

• Interrupt 4 is COM 1, the first serial port.
• Interrupt 5 is LPT2, the second parallel printer port.
• Interrupt 7 is LPTl, the first parallel printer port.

One must check that devices like a clone or emulation

board do not use the same interrupt level as the one used on
the network board (e.g., IRQ2) . If this is the case, the

interrupt setting on the network board will have to be changed

(e.g., to IRQ7) so there is no conflict.

G. TSR (TERMINATE AND STAY RESIDENT) PROGRAMS

A TSR is a program designed to remain in the computer's
random access memory (RAM) at all times so that users can

activate it with a keystroke, even if another program is also

in memory. TSR programs should be utilized with caution when

using DOS because DOS does not operate in protected mode. DOS

has no provisions for keeping one program from invading the

55
memory space of another, and such invasions cripple the

invaded program or causes a crash.

TSRs can cause software conflicts that masquerade as

hardware conflicts. They are often loaded via the AUTOEXEC.BAT

file, or may be loaded later on, as needed. The problem is

that once a TSR terminates, it does not release its space for

other purposes. This exacerbates the problem of limited RAM

that is available when PCs are used on a network (due to the

fact that both DOS and NOS must be resident in each PC) . One

solution is to use a "mark-release" approach. A "mark"

program is executed prior to loading the TSR to mark the place
in RAM where the TSR begins. A "release" program is executed
when the TSR terminates to release the RAM occupied by the

TSR.

56
 CONCLUSION

The LAN industry has come a long way in finding ways to

resolve compatibility issues. When local area networks first

came about, compatibility did not exist. There was no such

thing as multi-vendor networking or interoperability. Today

multi-vendor networking has improved. The improvement has

occurred mostly in inter-LAN communication by using protocols

like TCP/IP and routers and gateways. Unfortunately, many

compatibility problems persist in intra-LAN communication

among network boards, PC, NOS, DOS, and application software,

particularly when multiple vendors are involved. The most the

LAN manager can do is to be aware of potential

incompatibilities and to defend against them using the methods

suggested in this thesis.

57
APPENDIX: AS/IS LAB MANAGER'S GUIDE

58
 A Basic Reference Guide

For use on the

Administrative Sciences/Information Systems

(AS/IS) Computer Laboratories In

1-158, 1-224, and 1-250

September 1991

59
 TABLE OF CONTENTS

A. INTRODUCTION 4

B. NETWORK TECHNOLOGY 4

1. EtherNet 4

2 IBM Token-Ring 6

3. Appletalk 7

C. LABORATORY SETUP 8

1. 1-158 8

2. 1-224 8

3. 1-250 9

D. FILE ORGANIZATION 17

E. SETTING UP DIRECTORIES 17

F. SOFTWARE INSTALLATION 19

G. BATCH PROCESSING 19

1. Using Batch Files to Simplify Procedures . . 20

2. The CONFIG.SYS File 21

3. The AUTOEXEC.BAT File 22

4. DOS Environment 23

60
A. INTRODUCTION

The Administrative Sciences (AS) Department of the Naval

Postgraduate School has three microcomputer laboratories

They are located in Ingersoll Hall, rooms 1-158, 1-224, and I-

250. Two of these labs (1-224 and 1-250) are used as

instructional laboratories. Access to the labs are restricted

to students, faculty, and staff of the Naval Postgraduate

School

The purpose of this manual is to provide basic information

about the setup and concepts governing the Administrative

Sciences/Information Systems (AS/IS) microcomputer labs so as

to aid lab managers. It is not a step-by-step guide in

learning how to use these networks but rather an outline of

the types of networks available, their setup, physically and

logically; their capabilities; and the hardware and software

used in each network. The guide is intended for use only at

the Naval Postgraduate School and is tailored to the specific

configurations of these labs.

B. NETWORK TECHNOLOGY
There are three different technologies used in the AS/IS

laboratories: IBM Token-Ring, EtherNet, and Appletalk.

61
 1 . EtherNet

Ethernet Networks are defined by the IEEE 802.3

specification but, in practice, many variations also exist.

EtherNet networks have two points in common: a contention

access scheme and a linear bus topology. Ethernet runs on

coaxial cabling using baseband or broadband transmissions.

Ethernet can also be set up using telephone type cabling or

optical fiber but the system may not entirely conform to the

IEEE standard.

The original EtherNet networking scheme called for a

thick coaxial cable, which has come to be known as thick

EtherNet cable. Following the classic bus topology (a one-

cable network tapped at any point where a workstation is

required) , a single thick EtherNet trunk cable runs from one

end of the network to the other. The stations are connected

to the bus via transceivers and associated cabling.

Most PC-based EtherNet networks use the thinner RG-58

coaxial cable, and have been dubbed "Cheapernet" . Stations

are connected on a bus in which cable is strung from one

station to the next. A T-connector is used to attach the

cable to the network interface board in each network station.

EtherNet uses the CSMA/CD medium access method. If a

workstation wants to transmit, and no one else is using the

cable, the station begins transmitting. If the cable is in

62
use, the transmission is delayed until the cable is free. It

provides a transmission speed of 10 megabits per second

(Mbps) . (EtherNet implementations range from 1Mbps to 10Mbps,

although the 802.3 specification is intended to support signal

rates up to 20 Mbps)

2 . IBM Token-Ring

Token-Ring networks are defined by the IEEE 802.5

specification. Its name implies a ring, however, the token

ring is physically a star, and electrically a ring. The LAN

hardware in a Token-Ring network includes the network

interface card (NIC) , also called the network adapter; the

data cable; the patch or adapter cable; and the Multistation

Access Unit (MAU) , which is a wiring concentrator that can

connect as many as eight stations into the network. Repeaters

can be added between two Multistation Access Units (MAU's) to

increase the transmission distance. Each node acts as a

repeater—receiving the serial bit stream from the nearest

active upstream node, processing as necessary, and sending the

bit 3tream on down the cable to the next node in line. Only

a few bit times of delay are required in each network station

for these functions. The serial transmission follows a

complete ring or loop, with the sending station eventually

receiving its transmitted information back after this

information completes one round trip around the ring.

63
 These networks use a token-passing scheme for network

access. A station that needs to send a transmission must

first gain control of the token. The token is merely a

specific bit sequence that circulates among the nodes. For

this reason, these networks are often described as distributed

polling environments. Each workstation is thus polled to

determine if it needs access to the network. When a node is

in possession of the token, it can transmit a message that is

in its output buffer. Otherwise, the node is in bit-repeat

mode.

The wiring plans are star-wired rings, and shielded

twisted pair wiring is the standard for internode connection.

Current Token-Ring implementations are available in 4Mbps and

16Mbps.

3 . Appletalk
Appletalk is a baseband network with a bus topology.
It operates over shielded twisted-pair wire, unshielded

twisted-pair, or optical fiber--all at 230.4 kilobits per

second. Appletalk is a proprietary network standard that does

not conform to any IEEE specifications. But it uses a layered

architecture, and its protocols at every layer have been

published, so other vendors can develop products for it.

Appletalk networking uses an access scheme that is

similar to EtherNet's CSMA/CD — CSMA/CA. Under the CSMA/CA

64
scheme, each MAC on the network listens for the signal that

indicates another station is transmitting. If no other

station is transmitting (no signals for 400 microseconds) a

sending station waits an additional, random amount of time

before transmitting. The sending and receiving stations will

maintain control of the network medium by not allowing the

medium to be quiet for more than 200 microseconds.

C. LABORATORY SETUP

The following setup describes the contents of the AS/IS

labs. The table following the lab descriptions outlines the

different characteristics of the AS/IS networks.

1. 1-158

1-158 has six computers connected in an IBM Token-Ring

network. This network has six IBM-XT computers, each with

accelerator (Turbo) boards for increased speed. An IBM- AT is

used as a file and print server. The printer is a laser-

quality IBM 3812 Pageprinter. A Polaroid Palette film

recording system is attached to one user computer which allows

the creation of 35mm color slides.

Another, standalone, IBM-XT has the MAESTRO system,

which gives users the capability to create computer-aided

teaching programs. 1-158 also has a Unix workstation--the

Symmetric 375. See Figure (1) for a detailed layout.

65
 2. 1-224

This lab contains three networks: An IBM Token-Ring

network with 15 attached user computers (12 AT-clones, 3 IBM-

XTs) ; a 3COM EtherNet network with five attached computers;

and an Apple-TOPS Appletalk Network with seven computers

attached (six Apple Macintosh Plus' and two IBM-XTs) . This

Apple Talk network has APPLE- IBM connectivity.

The IBM Token-Ring network has two generic 38 6S and

one IBM-XT computers as file and printer servers. The XT also

serves as the IBM PC 3270 Emulation Gateway for 10 of the 15

user computers which are configured for terminal emulation.

In addition, a 40MB Bernoulli Box is attached to one of the

386 servers. There are also an AST optical scanner, two IBM

Proprinters, and one IBM Color Inkjet printer. See Figure (2)

for a detailed layout.

The 3COM EtherNet network uses a 3COM 3Server 3 Server,

an IBM Proprinter, and an IBM color plotter. See Figure (3)

for a detailed layout.

The AppleTalk network uses an Apple Macintosh Plus as

a server, which is located in a locked, vented cabinet. Apple

does not provide any inherent physical security to the server

unlike the IBM networks which allow locking of the keyboards.

The server software used is the server version of Appleshare.

In addition to the TOPS protocol, (which allows MAC-to-PC

66
communication) , the Apple network uses its own Apple protocol.

The printer in the network is an Apple LaserWriter. Its

cables and connectors are unique in that the network uses only

Appletalk cables and connectors. See Figure (4) for a

detailed layout.

3. 1-250

1-250 has an IBM PC (Broadband EtherNet) network with

28 attached computers (25 IBM-XT user computers and three IBM-

AT file and printer servers) . It also has two IBM

Proprinters, one IBM PC Graphics printer, one ceiling mounted

projector for displaying the instructor' s computer screen and

a dual-drive Bernoulli Box with a capacity of 40Mb. See

Figure (5) for a detailed layout. See below for a table

outlining the characteristics in the AS/IS network labs.

AS/IS Network Labs Table: (Schneidewind, 1991)

Characteristics IBM PC NET IBM Token-Ring 3COM Appletalk

EtherNet

(1-250) (1-224) (1-224) (1-224)

IEEE Standard 802.3 802.5 802.3 none

Aactu Method - CSMA/CD Token passing same aa IBM PC CSMA/CA

- Collision NET
detection
— Jam signal

67
 Medio Type Coax cable Shielded ooax cable Sheilded t

Twiated pair Unshielded

Twisted
pair

Trammijiion Broadband Baseband Baseband Baseband

Mat hod

Tranamiaaion Spaed 2 MBPS 4 C 16 MBPS 10 MBPS 230.4 kbpa

Diatance Limitation 1000 ft - 770 m. for STP - 1000 ft for 300 meters

between nodea and thin EtherNet bet nodea

.

repeaters - 3280 ft for

- 2 Km. for fiber thick EtherNEt

optica bet. MAU &

optical coupler

Max. Number of 72 - 72 for 4 Mbps 100 for thin 32

nodes - 250 for 16 Mbps EtherNet cable

Recommended Number 32 32 with 4Mbps 10 for thin 6-10

of nodea EtherNet cable

Message Format - Server same aa IBM PC EtherNEt packets Appleshara

message block net meaaagea
— network
message

User Interface - Dir. of same as IBM PC same aa IBM PC Appleahare

application BAT Net Net hard disk

filaa icons
- Network

Function Manua
- Network

Commands

Performance good performance - High overhead good performance low power,

at low loads at low load at low loads slow

— good

performance at
medium and high
loads

68
 Topology - logical bus - logical ring - logical bua -logical
- physical atar - phyaical atar - phyaical bua bua
- phyaical

bua

Othar Strvioi - a-mail aama am IBM PC - a-mail IBM

- limitad fila Nat - manu ayatan connacti-

tranafar ( PC to vity uaing

PC) TOPS
- full fila

tranafar via

dial— up

communucat ion

with hoata and

3270 annul.

D. FILE ORGANIZATION

System and application programs are installed by the

network supervisor or lab manager. A network administrator

must identify who uses certain files and what type of use the

files require. Although a great deal of data may be stored on

the network in potentially shareable form, opening up all the

data files to every user is not necessary or even desirable.

Users have different uses for the files. One user may need to

see a file but may not be involved in updating the information

the file contains. This type of user would be given Read-Only

(R-0) privileges. A user who actually updates the same file

would require Read/Write (R/W) privileges. In addition, this

kind of access restriction improves the integrity and the

69
 IN 158
IBM Token Rinp LAN
In back of 1-158

Print & File

SERVER
(IBM AT)

TN4
I J I

IBM
3812
Pageprinter

TN33
(MODEM)
TN6 TN7
(MODEM) (MODEM)

TN7M
(MODEM)
TN11
(MODEM)

Legend:

D
IBM 8228
MULTISTATION
U ACCESS
u UNIT
D (MAU)

70
 IN 224
IBM Token Ring LAN

(Future campus
3174-1L Bernoulli
backbone connection)
Box
(Mainframe
Controller) L^ ZZ3

3270 /printfrN Printer & /printerN Printer & INSTRUCTOR

COMPUTER
Gateway
SERVER
V #1 J File
V #2 J File
SERVER (IBM XT)
SERVER
(IBM XT) GBM XT) (IBM AT) TNI 8
(MODEM)
TNO f COLOR TN6M TN3
(3270)
JETPRINTER

J
IBM 8228
MULTISTATION
ACCESS
UNIT
(MAU)
I
AT AT AT AT AT AT
CLONE CLONE CLONE CLONE CLONE CLONE
DOCUMENT
TN20 TN21 TN22 TN23 TN24 TN25 (GRAPHIC)
(MODEM) (MODEM) (MODEM) (3270) (MODEM) (MODEM) SCANNER

IBM 8228
MULTISTATION
ACCESS
UNIT
(MAU)

TN26 TN27 TN28 TN29 TN30 TN31

(3270) (3270) (3270) (3270) (3270) (3270)

IBM 8228
MULTISTATION
ACCESS
UNIT
(MAU)
TNT 5
(3270)

Note: This drawing, not to scale, shows the layout of the TR-LAN.
*The 3270 Emulation connection to the mainframe goes thru the Gateway server.
* Uses Printer #1 ... ), ^ r Jel nnter
.
T ,
.
, ,

A Uses Printer #2 userS Can USC P )

y
71
 IN 224
3COM ETHERNET LAN

TN9 TN14 TN19 TN10

(ENET1) (ENET2) (ENET3) (ENET4)

NPS
TCP/IP ETHERNET
BACKBONE SINGLEPORT
INTERNET REPEATER
CONNECTION

2
3COM IBM
( Printer \ Server COLOR PLOTTER
(3Server3)

TN13
(ENET5)
NORMS
Legend:

=s33]
fl
Terminator 'T" Connector

72
 IN 224
Apple AppleTalk Layout

Secure
|"se. Cabinet

Macintosh
Plus Macintosh Macintosh Macintosh Macintosh Macintosh Laser
(Server) Plus Plus Plus Plus Plus Writer
45 Plus
Hard Disk

rTl
/£±
TOPS TOPS
IBM
XT
TN16 TN8.M

LEGEND:
-D=
LocalTalk Connector Kit

73
 IN 250
Ethernet LAN
(Broadband)

PRINTER N29 N30 N31 N32 N33

LEGEND: pa
'_••..•..
3gg§

m
Short Distance Kits Base Expander
*<&*
Connection Hardware

* The 3270 Emulation connects directly to the mainframe.

74
security of the data.

Users of the AS/IS labs are generally given Read-Only

privileges to application programs, which enables them to load

and execute the programs, but not modify them. In these labs,

however, the network supervisor has all rights to system files

and keeps documentation as to file access rights.

E. SETTING UP DIRECTORIES

The structure of the AS/IS labs network directories is

hierarchical. Each data storage area, or directory, is

logically "below" another directory. The directory that all

directories are below is called the root directory,

represented with a backslash character (\) . All other lower

level directories are called subdirectories.

The user computer operating system is usually placed in

its own directory, called DOS, below the root directory.

The objective for organizing a network is to give each

user only as much authority as is needed to fulfill a

particular job. In the AS/IS labs, all application programs

are stored on the server hard disks where they are most easily

controlled and supported and where they can be shared with all
users. To store their created files, users save them on their

own formatted 5.25 floppies.

75
 When users log onto the network, the first thing they see

is a main menu listing the different application programs and

files, from which they can make a choice. This approach

relieves the network manager from teaching users about paths

and directories. Explicit instructions are posted beside each

user computer on how to logon to the network. Users are not

required to know anything at all about the network. Menu

choices made by users are usually controlled by a batch file

that contains a number of instructions for creating the

necessary paths and loading the application. The logon

process occurs automatically without the user seeing anything

more than the options and the results.

F. SOFTWARE INSTALLATION

The network supervisor/lab manager ensures that the

appropriate version (s) of software and related utilities are

available to the users. Users will normally only use one

version of a particular application. This ensures that all

files created with the application will be compatible. User

training and support are then simplified because they need to

cover only one version of the software.

G. BATCH PROCESSING
No single tool is more important to the network user and

supervisor than the batch file. They are crucial in managing

76
and simplifying the network environment.

The AS/IS labs are controlled by batch files. A batch

file is a text file containing a series of operating system

commands. These files store the commands in text files and are

identified by the extension BAT. The commands in the batch

file execute sequentially when called upon. You can invoke

several, even hundreds of instructions by simply starting the

batch file with a one-word command. In the AS/IS labs, a

batch file is installed that starts the network and takes you

to a main menu where you can make a selection as to the

application you want to use. The selections on the main menu

are application batch files. The user can work with the

application and is completely free to use any of its features.

The AS/IS labs batch files are written so that when the user

exits the application, additional system commands are executed

to return the user computer to a standard network

configuration. He or she is then returned to the main menu

for another selection.

1 . Using Batch Files to Simplify Procedures

Batch files are versatile and simplify procedures by

carrying out repetitive sequences of commands. The user does

not have to try to remember all the necessary commands. For

77
example, if you frequently issue the same five commands when

you start a system, you can put those commands in a batch file

and give the batch file a name: for example STARTUP.BAT. When

you enter the command STARTUP, the file runs and the commands

are executed.

Batch files do not have to be complex chunks of

programming. Most batch files are short lists of operating

system commands. A typical batch file is no more than a dozen

lines

To create effective batch file systems, you must have

an understanding of the user computer' s operating system and

the network's operating system.

The Disk Operating System (DOS) used on the networks

is IBM version 3.3 except for the 3COM network in 1-224 which

uses IBM DOS 3.2. The Networking Operating System (NOS) used

in the AS/IS labs is version 1.2 of the IBM PC LAN program.

To create a Batch File, you can use one of three

different methods:

1 . The COPY Command - The Copy command is quick and easy

to use but is unsuitable for lengthy batch files because the

command does not enable you to make corrections without

completely re-creating the file. You can use COPY to copy

keyboard input into a file. You must ensure, however, that

78
the file name you specify does not already exist. Otherwise,

the old file is overwritten and lost.

2. The EDLIN program - This DOS utility supports the

creation and modification of batch files. This method is

always available because it is included with DOS.

3. The Word Processor - This highly flexible method is

ideal for creating and documenting long batch files. Load your

word processor and open a document. Type each command at the

left margin on a separate line. When you are finished, save

the file as an ASCII (or DOS text) file. Use any filename

you want, but be sure to use the BAT extension required for

all batch files.

2. The CONFIG.SYS File

The CONFIG.SYS file is not a batch file but rather a

system file that performs specific tasks. The file contains

values that DOS uses to configure itself in memory each time

the system starts. Such values include, for example, the

BUFFERS= entry (which defines the number of disk buffers DOS

uses) and the FILES= entry (which specifies the number of

files DOS can have open at one time) . The CONFIG.SYS file is

usually created the first time you install DOS. Later on, you

may modify the file to suit your changing requirements. The

file can include a DEVICE line, which identifies any one of a

number of physical or logical attachments to the network.

79
Each such device requires its own device driver, which is a

special program providing the required logical interface

between the device and the entire system. DOS must be able to

find the driver in order to load it, and the DEVICE line

specifies the path (if any) to the driver.

Each time the computer is turned on, DOS installs

itself in the computer's memory before executing any other

commands. DOS uses the entries in the CONFIG.SYS file to

customize the operating system. If this file does not exist,

DOS uses default values. Only after DOS is completely

installed does it check for the existence of the AUTOEXEC.BAT

file.

3. The AUTOEXEC.BAT File

When DOS starts, it searches the root directory for a

file calledAUTOEXEC.BAT. This distinctive file AUTOmatically

EXECutes a group of initializing instructions before doing

anything else. This BATch file is similar to other batch

files except that its filename is known to DOS and thus cannot

be U3ed for any other purpose except to startup a computer.

In the AS/IS labs, the AUTOEXEC.BAT sets up paths to

the drives and directories, loads certain operating files

(when needed) , and establishes environmental variables, as

required. It also sets parameters so that the prompt shows

the current directory. Finally, it displays a screen to give

80
users some information on starting the network, availability

of printers, and what to do if a problem arises.

Using a CONFIG.SYS file and an AUTOEXEC.BAT file will

go a long way toward relieving the user of having to wade

through a long series of complex instructions every time the

system is turned on. They also help avoid errors that might

occur during the start up process.

4 . DOS Environment

The DOS environment or Master DOS Environment Block is

a memory block in which the system stores critical information

created by the CONFIG.SYS and the AUTOEXEC.BAT files, such as

the command path and definition of the system prompt. To view

the current environment, type SET at the DOS prompt. The SET

command changes or displays the value assigned to an

environment variable. You can also use it to define

replaceable parameters for use with batch files. For example,

if the current environment space is too small, you can set

aside more space for the environment by editing the CONFIG.SYS

file.

When considering the potential sources of network

problems, it is important to remember that a LAN environment

includes not only the physical components such as hardware,

software and cabling but also the users of the LAN. Users may

often have difficulty with a network because of an incomplete

81
understanding of basic LAN concepts. A well organized network
administration avoids most of these problems by using

standardized procedures and by providing information and help

to the users.

Standard maintenance procedures help reduce a great number

of potential network problems. LAN maintenance involves not

only fixing or replacing broken parts, but includes performing

regular (and sometimes menial) tasks to reduce the risk of

network downtime. Simple procedures such as cleaning the

heads on a tape back-up unit, hard disk optimization routines,

periodic cabling inspections, and vacuuming and cleaning

network components are examples of tasks which may help to

prevent a large number of potential problems.

Cabling problems include network data and equipment power

cables. They can be wiggled or jarred loose, and become

disconnected, to cause computer or printer failures. Checking

the tightness of all cable connections is a sound

troubleshooting/maintenance action
Software problems can result from improper installation or

incorrect parameter settings for network services. Changes in

parameter settings should only be made when a clear

understanding of the desired effects are known! Software

problems can also occur when certain applications are

introduced to the network. Single user applications differ

82
significantly from programs which are network aware or are

specifically designed to be multiuser. Always thoroughly

check the documentation to understand the potential

implications. Careful testing should be performed before new

applications are made available to all network users!

83
 REFERENCES

Aaron, Bud and Aaron, Alex, Upgrading PC's Made Easy , Osborne
McGraw-Hill, 1991.

Apple Computer Inc., Understanding Computer Networks, Addison-

Wesley Publishing Company, 1989.
Archer, Rowland, The Practical Guide to Local Area Networks,
Osborne McGraw-Hill, 1986.
Axelrod, C. Warren, Computer Productivity , John Wiley & Sons, Inc.,
1982.

Berry, Paul, Operating The IBM PC Networks, Sybex, 1986.

Bertsekas, Dmitri and Gallager, Robert, Data Networks , Prentice-

Hall, Inc., 1987.

Brenner, Aaron, "Getting Better," p. 14 LAN Magazine, March 1989.

Bulettw, Greg and Chacon, Michael, Understanding 3 COM Networks,

Wordware Publishing, Inc. 1991.

BYTE Magazine, June 1990. "Growing Pains."

Calingaert, Peter, Operating System Elements, A User Perspective,

Prentice-Hall, 1982.

Cashin, Larry, "Choosing a LAN OS: Proceed with Caution," p. 59,

Software Magazine, M arch, 1990.

Cole, Gerald D., Computer Networking for Systems Programmers, John

Wiley and Sons, Inc., 1990.
Costa, Linda C. and Glasgow, Mary R. "Good Software Demands Good
,

Documentation", IEEE Professional Communication Society Conference

Record IEEE Service Center.
,

Davis, George R. , The Local Network Handbook, McGraw Hill

Publications, 1982.

Derfler Jr., Frank J., PC Magazine Guide to Connectivity, Ziff-

Davis Press, 1991.

84
Doll, Dixon, "The Year Ahead: Changes and Challenges in
Networking," Network World, 24 December, 1990.

Durr, Michael and Gibbs, Mark, Networking Personal Computers, 3rd

Edition Que Corporation, 1989.
,

Edelhart, Mike, Novell NETWARE Power Tools, Bantam Books, 1991.

Foster, Jennifer, The Hardware Guide for Novell NetWare, John Wiley
and Sons, Inc., 1991.

Gookin, Dan, The ABC s of Upgrading Your PC, Sybex Inc., 1990.

Green, Jr. Paul E., Network Interconnection and Protocol

Conversion, IEEE Press, 1988.

Guengerich, Steven, "Interoperability is Key to LAN Operating

Systems," p.l, Network World, 1 October 1990.
Hader, Michael L., NETWARE Troubleshooting, Addison-Wesley
Publishing Co. Inc., 1991.
Haugdahl, J. S., Inside NetBIOS, 3rd Edition, Architecture
Technology Corporation, 1990.

Johnston-Turner, Mary, "Standards are Moving Targets for Harried

Users," p. 35, Network World Nov 19, 1990.
,

Jordan, Larry and Churchill, Bruce, Communications and Networking

for the IBM PC and Compatibles, 3rd Edition, Simon and Schuster,
1990.

Kibirige, Harry M. "Local Area Networks in an Online Information

,

Retrieval Environment, " National Online Meeting, Proceedings of the

Ninth National Online Meeting, New York, 10-12 May, 1988.
LAN Magazine, December 1988. "LAN Interface Cards."

Liebing, Edward and Neff Ken, NETWARE Server, Troubleshooting and

,

Maintenance Handbook, McGraw-Hill Inc., 1990.

Liebing, Edward, Knutson, Charles D., and Day, Michael, NETWARE
Workstation, Troubleshooting and Maintenance Handbook McGraw-Hill,
,

Inc., 1990.

Madron, Thomas W. Local Area Networks,

, The Next Generation, John
Wiley and Sons, 1990.

85
Marney-Petix, Victoria C. Networking
, and Data Communications,
Reston Publishing Co., 1986.

Martin, James, Telecommunications and the Computer, P rentice-Hall,

Inc., 1990.

McConnell, John, Internetworking Computer Systems, Interconnecting

Networks and Systems, Prentice-Hall, 1988.

McNamara, John E., Local Area Networks , Digital Press, 1985.

Miller, Mark A. , LAN Troubleshooting Handbook, M & T Publishing,

Inc., 1989

Miller, Michael A., Internetworking, M & T Books 1991.

Needleman, Raphael, InfoWorld: Understanding Networks, Simon and

Schuster 1990.

Norton, Peter, Inside the IBM PC and PS/2, Simon and Schuster,
Inc., 1990.

Nunemacher, Greg, LAN Primer, M & T Books, 1990.

Pfaf fenberger, Bryan, Que's Computer User^s Dictionary, Que

Corporation, 1990.

Rimmer, Steven W. , Networking with TOPS, Sybex 1990.

Rinzler, Alan and Gancher, David, The Computerland Guide to Local

Area Networks, Computerland Books, 1990.
Rosch, Winn L., The Winn Rosch Hardware Bible, Simon and Schuster,
Inc., 1989.

Salamone, Salvatore, "Getting From EtherNet to Token-Ring by

Bridge: Beyond the LAN Boundary," p.l, Network World, 7 January,
1991.

Schatt, Stan, Linking LANS, A Micro Manage r' s Guide, Windcrest

Books, 1991.

Schneidewind, Norman, IS 3503 Class Notes, 1991.

Stallings, William, Local Networks, An Introduction, Macmillan

Publishing Co., 1984.

86
Stallings, William, Handbook of Computer Communications Standards,
Local Network Standards, Volume 2, Macmillan Publishing Co.y 1987.

Turner, Steven, "The Network Manager's Compendium of Standards,"

p.l, Network World, 15 April, 1991.

Van Kirk, Doug, "Troubleshoot your New Add-In Boards," p. 141, PC

Computing Magazine, July, 1991.

Woodward, Jeff, The ABC's of Novell Netware, Svbex 1989.

Woram, John, The PC Configuration Handbook, 2nd Edition, Bantam

Books, 1990.

87
 INITIAL DISTRIBUTION LIST

1 Defense Technical Information Center

Cameron Station
Alexandria, Virginia 22304-6145

2 Superintendent
Attn: Library, Code 0052
Naval Postgraduate School
Monterey, California 93943-5000

3 Commanding Officer
Attn: LCDR Rita V. Espiritu
NAVCOMTELSTA
PSC 458, Box 7
FPO AP 96451

4. Professor Norman F. Schneidewind

Administrative Sciences Department
Naval Postgraduate School
Monterey, California 93943-5000

5. Professor Myung W. Suh

Administrative Sciences Department
Naval Postgraduate School
Monterey, California 93943-5000

88
Thesis
E6885 Espiritu
c.l Local area' network (LAN)
compatibility issues.