The OSI Model

defiantneedlessNetworking and Communications

Oct 23, 2013 (3 years and 9 months ago)

159 views

Chapter 1
The OSI Model
Complete Reference/ Cisco: TCR / Hill / 9280-1 / Chapter 1
3
Complete Reference/ Cisco: TCR / Hill / 9280-1 / Chapter 1
P:\010Comp\CompRef8\280-1\ch01.vp
Wednesday, February 13, 2002 2:14:29 PM
Color profile: Generic CMYK printer profile
Composite Default screen
Complete Reference/ Cisco: TCR / Hill / 9280-1 / Chapter 1
T
he OSI (Open Systems Interconnection) model is a bit of an enigma.Originally
designed to allowvendor-independent protocols and to eliminate monolithic
protocol suites,the OSI model is actually rarely used for these purposes today.
However,it still has one very important use:it is one of the best tools available today
to describe and catalog the complex series of interactions that occur in networking.
Because most of the protocol suites in use now(such as TCP/IP) were designed using
a different model,many of the protocols in these suites don’t match exactly to the OSI
model,which causes a great deal of confusion.For instance,some books claimthat
Routing Information Protocol (RIP) resides at the network layer,while others claimit
resides at the application layer.The truth is,it doesn’t lie solely in either layer.The
protocol,like many others,has functions in both layers.The bottomline is,look at the OSI
model for what it is:a tool to teach and describe hownetwork operations take place.
For this book, the main purpose of knowing the OSI model is so that you can
understand which functions occur in a given device simply by being told in which
layer the device resides. For instance, if I tell you that physical (Media Access
Control—MAC) addressing takes place at layer 2 and logical (IP) addressing takes
place at layer 3, then you will instantly recognize that an Ethernet switch responsible
for filtering MAC (physical) addresses is primarily a layer 2 device. In addition, if I
were to tell you that a router performs path determination at layer 3, then you already
have a good idea of what a router does.
This is why we will spend some time on the OSI model here. This is also why you
should continue to read this chapter, even if you feel you know the OSI model. You
will need to fully understand it for the upcoming topics.
What Is a Packet?
The terms packet, datagram, frame, message,and segment all have essentially the same
meaning—they just exist at different layers of the OSI model. You can think of a packet
as a piece of mail. To send a piece of snail mail, you need a number of components
(see Figure 1-1):

Payload This component is the letter you are sending, say, a picture of your
newborn son for Uncle Joe.

Source address This component is the return address on a standard piece
of mail. This indicates that the message came from you, just in case there is a
problem delivering the letter.

Destination address This component is the address for Uncle Joe, so that the
letter can be delivered to the correct party.

A verification system This component is the stamp. It verifies that you have
gone through all of the proper channels and the letter is valid according to
United States Postal Service standards.
4
C i s c o:T h e C o m p l e t e R e f e r e n c e
Complete Reference/ Cisco: TCR / Hill / 9280-1 / Chapter 1
P:\010Comp\CompRef8\280-1\ch01.vp
Wednesday, February 13, 2002 2:14:30 PM
Color profile: Generic CMYK printer profile
Composite Default screen
Complete Reference/ Cisco: TCR / Hill / 9280-1 / Chapter 1
NETWORKINGBASICS
A packet is really no different. Let’s use an e-mail message as an example—see
Figure 1-2. The same information (plus a few other pieces, which we will cover later
in the chapter) is required:
 Payload This component is the data you are sending, say, an e-mail to Uncle
Joe announcing your newborn son.

Source address This component is the return address on your e-mail. It
indicates that the message came from you, just in case there is a problem
delivering the e-mail.

Destination address This component is the e-mail address for Uncle Joe, so
that the e-mail can be delivered correctly.

Verification system In the context of a packet, this component is some type of
error-checking system. In this case, we will use the frame check sequence (FCS).
The FCS is little more than a mathematical formula describing the makeup of a
packet. If the FCS computes correctly at the endpoint (Uncle Joe), then the data
within is expected to be valid and will be accepted. If it doesn’t compute
correctly, the message is discarded.
The following sections use the concept of a packet to illustrate how data travels
down the OSI model, across the wire, and back up the OSI model to arrive as a new
message in Uncle Joe’s inbox.
C h a p t e r 1:T h e O S I M o d e l
5
Complete Reference/ Cisco: TCR / Hill / 9280-1 / Chapter 1
Figure 1-1.Snail mail components
P:\010Comp\CompRef8\280-1\ch01.vp
Wednesday, February 13, 2002 2:14:30 PM
Color profile: Generic CMYK printer profile
Composite Default screen
OSI Model Basics
The OSI model is a layered approach to networking. Some of the layers may not even
be used in a given protocol implementation, but the OSI model is broken up so that any
networking function can be represented by one of the 7 layers. Table 1-1 describes the
layers,beginning with layer 7 and ending with layer 1.I amdescribing themin this order
because,in most cases,people tend to understand the model better if introduced in
this order.
Complete Reference/ Cisco: TCR / Hill / 9280-1 / Chapter 1
6
C i s c o:T h e C o m p l e t e R e f e r e n c e
Complete Reference/ Cisco: TCR / Hill / 9280-1 / Chapter 1
Figure 1-2.Basic packet components
Layer Function
Application (layer 7) This layer is responsible for communicating directly with
the application itself. This layer allows an application to
be written with very little networking code. Instead, the
application tells the application-layer protocol what it
needs, and it is the application layer’s responsibility to
translate this request into something the protocol suite
can understand.
Presentation (layer 6) This layer is responsible for anything involved with
formatting of a packet: compression, encryption,
decoding, and character mapping. If you receive
an e-mail, for instance, and the text is gobbledygook,
you have a presentation-layer problem.
Table 1-1.The Layers of the OSI Model
P:\010Comp\CompRef8\280-1\ch01.vp
Wednesday, February 13, 2002 2:14:31 PM
Color profile: Generic CMYK printer profile
Composite Default screen
When data is sent from one host to another on a network, it passes from the
application; down through the model; across the media (generally copper cable) as an
electrical or optical signal, representing individual 0’s and 1’s; and then up through the
model at the other side. As this happens, each layer that has an applicable protocol
adds a header to the packet, which identifies how that specific protocol should process
the packet on the other side. This process is called encapsulation.See Figure 1-3 for a
diagram (note that AH stands for application header, PH stands for presentation
C h a p t e r 1:T h e O S I M o d e l
7
Complete Reference/ Cisco: TCR / Hill / 9280-1 / Chapter 1
NETWORKINGBASICS
Layer Function
Session (layer 5) This layer is responsible for establishing connections, or
sessions, between two endpoints (usually applications).
It makes sure that the application on the other end has
the correct parameters set up to establish bidirectional
communication with the source application.
Transport (layer 4) This layer provides communication between one
application program and another. Depending on
the protocol, it may be responsible for error detection
and recovery, transport-layer session establishment
and termination, multiplexing, fragmentation, and
flow control.
Network (layer 3) This layer is primarily responsible for logical addressing
and path determination, or routing, between logical
address groupings.
Datalink (layer 2) This layer is responsible for physical addressing and
network interface card (NIC) control. Depending on the
protocol, this layer may perform flow control as well.
This layer also adds the FCS, giving it some ability to
detect errors.
Physical (layer 1) The simplest of all layers, this layer merely deals with
physical characteristics of a network connection: cabling,
connectors, and anything else purely physical. This layer
is also responsible for the conversion of bits and bytes
(1’s and 0’s) to a physical representation (electrical
impulses, waves, or optical signals) and back to bits on
the receiving side.
Table 1-1.The Layers of the OSI Model (continued)
P:\010Comp\CompRef8\280-1\ch01.vp
Wednesday, February 13, 2002 2:14:31 PM
Color profile: Generic CMYK printer profile
Composite Default screen
8
C i s c o:T h e C o m p l e t e R e f e r e n c e
Complete Reference/ Cisco: TCR / Hill / 9280-1 / Chapter 1
header, and so on). Upon arriving at the destination, the packet will be passed back up
the model, with the protocol headers being removed along the way. By the time the
packet reaches the application, all that remains is the data, or payload.
Now we will delve into the specifics of each layer and the additional processes for
which each layer is responsible.
Layer 7: The Application Layer
The application layer is responsible for interacting with your actual user application.Note
that it is not (generally) the user application itself,but,rather,the network applications used
by the user application.For instance,in web browsing,your user application is your
browser software,such as Microsoft Internet Explorer.However,the network application
being used in this case is HTTP,which is also used by a number of other user applications
(such as Netscape Navigator).Generally,I tell my students that the application layer is
responsible for the initial packet creation;so if a protocol seems to create packets out of thin
air, it is generally an application- layer protocol. While this is not always the case (some
protocols that exist in other layers create their own packets), it’s not bad as a general
guideline.Some common application-layer protocols are HTTP,FTP,Telnet,TFTP,SMTP,
POP3,SQL,and IMAP.See Chapter 5 for more details about HTTP,FTP,SMTP,and POP3.
Figure 1-3.Data encapsulation as data is passed through the model
P:\010Comp\CompRef8\280-1\ch01.vp
Wednesday, February 13, 2002 2:14:33 PM
Color profile: Generic CMYK printer profile
Composite Default screen
Complete Reference/ Cisco: TCR / Hill / 9280-1 / Chapter 1
NETWORKINGBASICS
Layer 6: The Presentation Layer
The presentation layer is one of the easiest layers to understand because you can easily
see its effects.The presentation layer modifies the format of the data.For instance,I
might send you an e-mail message including an attached image.Simple Mail Transport
Protocol (SMTP) cannot support anything beyond plain text (7-bit ASCII characters).To
support the use of this image,your application needs a presentation-layer protocol to
convert the image to plain text (in this case,Multi-purpose Internet Mail Extensions,or
MIME).This protocol will also be responsible for converting the text back into an image
at the final destination.If it did not,the body of your message would appear like this:
BCNHS ^%CNE (37NC UHD^Y 3cNDI U&">{ }|__D Iwifd YYYTY TBVBC
This is definitely not a picture, and is obviously a problem, proving my point that a
presentation-layer problem is generally easy to recognize. The presentation layer is also
responsible for compression and encryption, and pretty much anything else (such as
terminal emulation) that modifies the formatting of the data.Some common presentation-
layer data formats include ASCII,JPEG,MPEG,and GIF.
Layer 5: The Session Layer
Conversely, the session layer is one of the most difficult layers to understand. It is
responsible for establishing, maintaining, and terminating sessions. This is a bit of a
broad and ambiguous description, however, because several layers actually perform
the function of establishing, maintaining, and terminating sessions on some level.
The best way to think of the session layer is that it performs this function between two
applications. However, as we will see in Chapter 5, in TCP/IP, the transport layer
generally performs this function, so this isn’t always the case. Some common session-
layer protocols are RPC, LDAP, and NetBIOS Session Service.
Layer 4: The Transport Layer
The transport layer performs a number of functions, the most important of which are
error checking, error recovery, and flow control. The transport layer is responsible
for reliable internetwork data transport services that are transparent to upper-layer
programs. The first step in understanding transport-layer error checking and recovery
functions is to understand the difference between connection-based and connectionless
communication.
Connection-Based and Connectionless Communication
Connection-based communication is so named because it involves establishing
a connection between two hosts before any user data is sent. This ensures that
bidirectional communication can occur. In other words, the transport-layer protocol
sends packets to the destination specifically to let the other end know that data is
coming. The destination then sends a packet back to the source specifically to let
C h a p t e r 1:T h e O S I M o d e l
9
P:\010Comp\CompRef8\280-1\ch01.vp
Wednesday, February 13, 2002 2:14:33 PM
Color profile: Generic CMYK printer profile
Composite Default screen
the source know that it received the “notification” message. In this way, both sides are
assured that communication can occur.
In most cases, connection-based communication also means guaranteed delivery.
In other words, if you send a packet to a remote host and an error occurs, then either
the transport layer will resend the packet, or the sender will be notified of the packet’s
failed delivery.
Connectionless communication,on the other hand, is exactly the opposite: no initial
connection is established. In most cases (although not all), no error recovery exists. An
application, or a protocol above or below the transport layer, must fend for itself for
error recovery. I generally like to call connectionless communication “fire and forget.”
Basically, the transport layer fires out the packet and forgets about it.
In most cases, the difference between connection-based and connectionless
protocols is very simple. You can think of it like the difference between standard
mail and certified mail. With standard mail, you send off your message and hope it
gets there. You have no way of knowing whether the message was received. This is
connectionless communication. With certified mail, on the other hand, your message
is either delivered correctly and you get a receipt, or your message is attempted to be
delivered many times before it times out and the postal service gives up—and you still
get a receipt. Either way, you are guaranteed to be notified of what happened so that
you can take appropriate measures. This is typical connection-based communication.
Flow Control
In it’s simplest form,flowcontrol is a method of making sure that an excessive amount
of data doesn’t overrun the end station.For example,imagine that PCAis running at
100 Mbps and PCB is running at 10 Mbps.If PCAsends something to PCB at full speed,
90 percent of the information will be lost because PCB cannot accept the information at
100 Mbps.This is the reason for flowcontrol.
Currently, flow control comes in three standard flavors, as described in the
following sections.
Buffering
Commonly used in conjunction with other methods of flowcontrol,
buffering is probably the simplest method.Think of a buffer as a sink.Imagine you have
a faucet that flows four gallons of water a minute,and you have a drain that accepts only
three gallons of water a minute.Assuming that the drain is on a flat countertop,what
happens to all of the excess water?That’s right,it spills onto the floor.This is the same
thing that happens with the bits fromPCAin our first example.The answer,as with
plumbing,is to add a “sink,” or buffer.However,this solution obviously leads to its own
problems.First,buffers aren’t infinite.While they work well for bursts of traffic,if you
have a continuous streamof excessive traffic,your sink space will eventually run out.
At this point,you are left with the same problem—bits falling on the floor.
Congestion Notification
Congestion notification is slightly more complex than
buffering, and it is typically used in conjunction with buffering to eliminate its major
Complete Reference/ Cisco: TCR / Hill / 9280-1 / Chapter 1
10
C i s c o:T h e C o m p l e t e R e f e r e n c e
P:\010Comp\CompRef8\280-1\ch01.vp
Wednesday, February 13, 2002 2:14:34 PM
Color profile: Generic CMYK printer profile
Composite Default screen
problems. With congestion notification, when a device’s buffers begin to fill (or it
notices excessive congestion through some other method), it sends a message to the
originating station basically saying “Slow down, pal!” When the buffers are in better
shape, it then relays another message stating that transmission can begin again. The
obvious problem with this situation is that in a string of intermediate devices (such as
routers), congestion notification just prolongs the agony by filling the buffers on every
router along the path.
For example, imagine Router A is sending packets to Router C through Router B
(as in Figure 1-4). As Router C’s buffer begins to fill, it sends a congestion notification
to Router B. This causes Router B’s buffer to fill up. Router B then sends a congestion
notification to Router A. This causes Router A’s buffer to fill, eventually leading to a
“spill” (unless, of course, the originating client understands congestion notifications
and stops the flow entirely). Eventually, Router C sends a restart message to Router B,
but by that time, packets will have already been lost.
Windowing
The most complex and flexible form of flow control, windowing, is
perhaps the most commonly used form of flow control today. In windowing,an agreed-
upon number of packets are allowed to be transferred before an acknowledgment from
C h a p t e r 1:T h e O S I M o d e l
11
Complete Reference/ Cisco: TCR / Hill / 9280-1 / Chapter 1
NETWORKINGBASICS
Figure 1-4.The problems with buffering and congestion notification
P:\010Comp\CompRef8\280-1\ch01.vp
Wednesday, February 13, 2002 2:14:35 PM
Color profile: Generic CMYK printer profile
Composite Default screen
12
C i s c o:T h e C o m p l e t e R e f e r e n c e
Complete Reference/ Cisco: TCR / Hill / 9280-1 / Chapter 1
the receiver is required. This means that one station should not be able to easily
overload another station: it must wait on the remote station to respond before sending
more data. In addition to flow control, windowing is also used for error recovery, as
we will see in Chapter 5.
Some common transport-layer protocols are TCP, UDP, and SPX, which will be
covered in more detail in Chapters 5 and 7.
Layer 3: The Network Layer
The network layer deals with logical addressing and path determination (routing).While
the methods used for logical addressing vary with the protocol suite used,the basic
principles remain the same.Network-layer addresses are used primarily for locating a
host geographically.This task is generally performed by splitting the address into two
parts:the group field and the host field.These fields together describe which host you
are,but within the context of the group you are in.This division allows each host to
concern itself only with other hosts in its group;and the division allows specialized
devices,called routers,to deal with getting packets fromone group to another.
Some common network-layer protocols are IP and IPX, which are covered in
Chapters 5 through 7.
Layer 2: The Datalink Layer
The datalink layer deals with arbitration, physical addressing, error detection, and
framing, as described in the following sections.
Arbitration
Arbitration simply means determining howto negotiate access to a single data channel
when multiple hosts are attempting to use it at the same time.In half-duplex baseband
transmissions,arbitration is required because only one device can be actively sending
an electrical signal at a time.If two devices attempt to access the mediumat the same
instant,then the signals fromeach device will interfere,causing a collision.This
phenomenon is perhaps better demonstrated in Figure 1-5.
Physical Addressing
All devices must have a physical address. In LAN technologies, this is normally a
MAC address. The physical address is designed to uniquely identify the device globally.
A MAC address (also known as an Ethernet address, LAN address, physical address,
hardware address, and many other names) is a 48-bit address usually written as
12 hexadecimal digits, such as 01-02-03-AB-CD-EF. The first six hexadecimal digits
identify the manufacturer of the device, and the last six represent the individual device
from that manufacturer. Figure 1-6 provides a breakdown of the MAC address. These
addresses were historically “burnt in,” making them permanent. However, in rare
cases, a MAC address is duplicated. Therefore, a great many network devices today
have configurable MAC addresses. One way or another, however, a physical address
of some type is a required component of a packet.
P:\010Comp\CompRef8\280-1\ch01.vp
Wednesday, February 13, 2002 2:14:35 PM
Color profile: Generic CMYK printer profile
Composite Default screen
Error Detection
Another datalink-layer function,error detection,determines whether problems with
a packet were introduced during transmission. It does this by introducing a trailer,
the FCS, before it sends the packet to the remote machine. This FCS uses a Cyclic
Redundancy Check (CRC) to generate a mathematical value and places this value in
the trailer of the packet. When the packet arrives at its destination, the FCS is examined
and the reverse of the original algorithm that created the FCS is applied. If the frame
was modified in any way, the FCS will not compute, and the frame will be discarded.
The FCS does not provide error recovery, just error detection. Error recovery is the
responsibility of a higher layer, generally the transport layer.
Framing
Framing is a term used to describe the organization of the elements in a packet
(or, in this case, a frame). To understand why this task is so important, we need to
C h a p t e r 1:T h e O S I M o d e l
13
Complete Reference/ Cisco: TCR / Hill / 9280-1 / Chapter 1
NETWORKINGBASICS
Figure 1-5.A collision and the resulting useless packet
Figure 1-6.Breakdown of a MAC address
P:\010Comp\CompRef8\280-1\ch01.vp
Wednesday, February 13, 2002 2:14:37 PM
Color profile: Generic CMYK printer profile
Composite Default screen
look at it from the device’s perspective. First, realize that everything traveling
over the cable is simply a representation of a 0 or a 1. So, if a device receives a string of
bits, such as 011010100010101111010111110101010100101000101010111, and so on, how
is it to know which part is the MAC address, or the data, or the FCS? It requires a key.
This is demonstrated in Figure 1-7.
Also, because different frame types exist, the datalink layers of both machines must
be using the same frame types to be able to tell what the packet actually contains.
Figure 1-8 shows an example of this.
Notice that the fields do not line up. This means that if one machine sends a packet
in the 802.3 format, but the other accepts only the Sub-Network Access Point (SNAP)
14
C i s c o:T h e C o m p l e t e R e f e r e n c e
Complete Reference/ Cisco: TCR / Hill / 9280-1 / Chapter 1
Figure 1-7.An Ethernet 802.3 framing key being applied to the bit stream, breaking
it into sections
Figure 1-8.Misaligned fields due to incorrect frame type
P:\010Comp\CompRef8\280-1\ch01.vp
Wednesday, February 13, 2002 2:14:39 PM
Color profile: Generic CMYK printer profile
Composite Default screen
format, they will not be able to understand each other because they are looking for
different components in different bytes of the packet.
Some common datalink-layer protocols are the following: virtually all of the 802
protocols (802.2, 802.3, 802.5, and so on), LAPB, LAPD, and LLC.
Layer 1: The Physical Layer
The physical layer is responsible for the most substantial of all functions.All connectors,
cabling,frequency specifications,distances,propagation-delay requirements,voltages—
in short, all things physical—reside at the physical layer.
Some common physical-layer protocols are EIA/TIA 568A and B, RS 232, 10BaseT,
10Base2, 10Base5, 100BaseT, and USB.
Peer Communication
Peer communication is the process in networking whereby each layer communicates
with its corresponding layer on the destination machine. Note that the layers do not
communicate directly, but the process is the same as if they were communicating
directly. A packet is sent from one host to another with all headers attached; but, as the
packet passes up through the model on the other side, each layer is solely responsible
for the information in its own header. It views everything else as data. This process is
shown in Figure 1-9.
C h a p t e r 1:T h e O S I M o d e l
15
Complete Reference/ Cisco: TCR / Hill / 9280-1 / Chapter 1
NETWORKINGBASICS
Figure 1-9.Peer communication
P:\010Comp\CompRef8\280-1\ch01.vp
Wednesday, February 13, 2002 2:14:40 PM
Color profile: Generic CMYK printer profile
Composite Default screen
16
C i s c o:T h e C o m p l e t e R e f e r e n c e
Complete Reference/ Cisco: TCR / Hill / 9280-1 / Chapter 1
Note that a layer is concerned only with the header from the exact same layer on
the other device. It treats everything else as data (even though it isn’t). Therefore, one
layer can, in a sense, communicate with its twin layer on the other device.
Bringing It All Together
Finally,I have included a sample network communication between two devices,broken
down by layer (see Figure 1-10).Note that this sample is not technically accurate.I have
included it only for illustrative purposes because it shows howeach layer performs a
specific function,even if that function isn’t performed in exactly the same manner in
real life.The major technical problemwith this diagramlies at the network layer,in the
“Intermediate Destination Address” field.There is no Intermediate Address field in
reality,but because we have not discussed howrouting really works yet,this example
illustrates the point well enough for now.
In this example,we are sending an e-mail using TCP/IP.As we transmit the message,
it begins at layer 7 by adding a Mail Application Programming Interface (MAPI) header.
Then it passes to the presentation layer,which adds a MIME header to explain the
message format to the other side.At the session layer,name resolution is performed,
resolving techtrain.comto 209.130.62.55.At the transport layer,the 256KB message is
segmented into four 64KB chunks,and a TCP session is established,using windowing
for flowcontrol.At the network layer,routing is performed,and the path is sent to the
nearest router (represented here by the Intermediate Destination Address).
Also note that the IP addresses (logical) are resolved to MAC addresses (physical)
so that they can be understood by the next layer. At the datalink layer, the packet is
segmented again, this time into frames that conform to the Maximum Transmission
Unit (MTU) of the media. At the physical layer, the data is sent as electrical signals. At
the other side, the communication passes back up through the model, performing the
opposite of the sending machine’s calculations to rebuild the packet into one 256KB
chunk of raw data for the application.
Other Network Models
The DODmodel is important because it is the foundation for TCP/IP,not the OSI model.
While the DODmodel matches the OSI model fairly well,the fact that it is the foundation
for TCP/IP can lead to some confusion when attempting to learn the OSI model.The
P:\010Comp\CompRef8\280-1\ch01.vp
Wednesday, February 13, 2002 2:14:41 PM
Color profile: Generic CMYK printer profile
Composite Default screen
C h a p t e r 1:T h e O S I M o d e l
17
Complete Reference/ Cisco: TCR / Hill / 9280-1 / Chapter 1
NETWORKINGBASICS
Figure 1-10.Processes performed by each layer of the model
P:\010Comp\CompRef8\280-1\ch01.vp
Wednesday, February 13, 2002 2:14:45 PM
Color profile: Generic CMYK printer profile
Composite Default screen
18
C i s c o:T h e C o m p l e t e R e f e r e n c e
Complete Reference/ Cisco: TCR / Hill / 9280-1 / Chapter 1
upper layers of the DODmodel don’t match the upper layers of the OSI model,which
can lead to different books listing protocols in different places within the OSI model.
The key here is to understand that unless you are studying for a test,it doesn’t really
matter too much where you place a given protocol in the OSI model,as long as you
understand the functionality of each layer of the model.Figure 1-11 depicts howthe
OSI and DODmodels match up.
Whereas the OSI and DOD models present a model of hownetwork-based
communication occurs,Cisco’s hierarchical internetworking model is a layered approach
to the topological design of an internetwork.It is designed to help improve performance,
while at the same time allowing optimumfault tolerance.When you use this model,
you simplify the network design by assigning various roles to the layers of the network
design.The obvious drawback of using this model in a small- to medium-sized network
is cost;however,if you require a high-performance,scalable,redundant internetwork,
using this approach is one of the best ways to design for it.
The hierarchical internetworking model consists of three layers:
 Core layer This layer is the network backbone. As such, the main issue here is
that any major problem will likely be felt by everyone in the internetwork. Also,
because speed is very important here (due to the sheer volume of traffic that
will be entering the backbone), few activities that consume significant routing
or switching resources should be applied in this layer. In other words, routing,
access lists, compression, encryption, and other resource-consuming activities
should be done before the packet arrives at the core.
 Distribution layer This layer is the middle ground between the core and
access layers. Clients will not be directly connected to this layer, but most of
their packet processing will be performed at this layer. This is the layer where
most supporting functions take place. Routing, Quality of Service (QoS), access
Figure 1-11.The DOD and OSI models
P:\010Comp\CompRef8\280-1\ch01.vp
Wednesday, February 13, 2002 2:14:45 PM
Color profile: Generic CMYK printer profile
Composite Default screen
C h a p t e r 1:T h e O S I M o d e l
19
Complete Reference/ Cisco: TCR / Hill / 9280-1 / Chapter 1
NETWORKINGBASICS
lists, encryption, compression, and network address translation (NAT) services
are performed at this layer.

Access layer This layer provides user access to local segments. The access
layer is characterized by LAN links, usually in a small-scale environment (like
a single building). Put simply, this layer is where the clients plug in. Ethernet
switching and other basic functions are generally performed here.
Figure 1-12 provides an example of the model in action.
Figure 1-12.The Cisco hierarchical internetworking model
P:\010Comp\CompRef8\280-1\ch01.vp
Wednesday, February 13, 2002 2:14:47 PM
Color profile: Generic CMYK printer profile
Composite Default screen
Summary
In this chapter, we have reviewed the most popular networking models, including the
OSI, Cisco, and DOD models. This information will help us understand references to
the layered networking approach examined throughout the book, and will serve as a
guide to understanding the place of routing and switching in any environment.
20
C i s c o:T h e C o m p l e t e R e f e r e n c e
Complete Reference/ Cisco: TCR / Hill / 9280-1 / Chapter 1
P:\010Comp\CompRef8\280-1\ch01.vp
Wednesday, February 13, 2002 2:14:47 PM
Color profile: Generic CMYK printer profile
Composite Default screen