Charles M. Kozierok The TCP-IP Guide

Подождите немного. Документ загружается.

In some areas, the layers are so closely related that the lines between them become blurry.

This is particularly the case when looking at the higher layers; many technologies

implement two or even all three of these layers, which is another reason why I feel they best

belong in a group together. One important reason why the distinctions between layers five

through seven are blurry is that the TCP/IP protocols are based on the TCP/IP model,

which combines the functions of layers five through seven in a single, thick layer.

Key Concept: The four lower layers of the OSI model are most often discussed

individually, because the boundaries between them are reasonably clear-cut. In

contrast, the lines between the session, presentation and application layers are

somewhat blurry. As a result, sometimes protocols span two or even all three of these

layers; this is especially true of TCP/IP application protocols, since the TCP/IP model treats

layers five through seven as a single layer.

Sublayers

Finally, note that some OSI Reference Model layers are further divided into sublayers to

help define more precisely the internal details of protocols and technologies at those layers.

This is commonly done at the lower layers, especially the physical layer and the data link

layer.

"N" Notation and Other OSI Model Layer Terminology

As a theoretical model, the OSI Reference Model comes complete with a set of terminology

that is used to describe it and its constituent parts. This is sort of both good news and bad.

The good news is that if you understand this terminology, it can help you comprehend much

better most OSI model discussions, and also how technologies relate to the model. The bad

news is that the terminology can also increase confusion—especially since it isn't always

used consistently.

Common OSI Model Terminology Concepts

Let’s discuss a few of the terminology concepts you will often see used to refer to the OSI

Reference Model. Many of these are illustrated in Figure 12.

Layer Names and Numbers

The various layers of the OSI Reference Model are referred to in a variety of ways. They

may have their names spelled out in full, or they may be abbreviated. They are also often

simply referenced by their layer number. So for example, all of these refer to the same

thing: “data link layer”, “Data Link Layer”, “DLL”, “L2”, “layer two”, “layer 2”.

Similarly, you will often see layer names being used as adjectives to describe protocols and

technologies. A “layer 3 technology” is one that operates primarily at the network layer.

“N” Notation

The letter “N” is often used to generically refer to a number within the computer world. With

respect to the OSI Reference Model, it's common to see this letter used in discussions that

relate generically to individual layers without mentioning a specific layer. You will hear terms

like “N-functions” and “N-services”, which just refer to the functions and services provided

within a particular layer.

As another example, you might see someone say that a particular technology “provides a

useful service to the N+1 layer”. This just means it provides a function to the layer above

the one at which it operates. Conceptually, every layer but the first and seventh have an “N-

1” layer and an “N+1” layer, and so on. If you are looking at the network layer (layer 3) then

the “N+2 layer” is the session layer (layer 5). See the discussion of model layer interfaces

for more information.

Figure 12: OSI Reference Model Layer Relationships and Terminology

This diagram shows the relationships between OSI model layers and the terminology used to refer to adjacent

layers in the context of any particular layer. Each layer (except layer 7) provides services the layer above it;

each layer (other than layer 1) uses services provided by the layer below. Another way of saying this is that

each layer “N” provides services to layer “N+1” and uses the services of layer “N-1”. Taking the example of

layer three, the network layer, we see that it provides services to layer four, and uses services of layer two.

From the standpoint of the network layer, the transport layer is layer “N+1” and the data link layer “N-1”.

N+4 Application

N+3 Presentation

N+2 Session

N+1 Transport

N Network

N-1 Data Link

N-2 Physical

Layer 3

Services

Provided

Layer 2

Services

Provided

Layer 1

Services

Provided

Layer 4

Services

Provided

Layer 5

Services

Provides

Layer 6

Services

Provided

Layer 3

Services

Used

Layer 2

Services

Used

Layer 1

Services

Used

Layer 4

Services

Used

Layer 5

Services

Used

Layer 6

Services

Used

Protocols and Interfaces

These words have special meaning within the context of the OSI model: a protocol repre-

sents communication between logical or physical devices at the same layer of the model

while an interface represents information moving between adjacent layers within the same

device.

Thus, in “N” notation, protocols represent communication between layer N on one device

and layer N on another device, while interfaces deal with communication between layer N

and N+1 or Layer N and N-1 on the same device. The topics on OSI protocols and OSI

interfaces describe these issues much better.

Network Stacks

What do you get when you take a bunch of layers and put them all one on top of each

other? Well, you get a stack. This term is used to refer to either the entire set of layers in a

model or suite of technologies, or a partial set. Since each layer has protocols associated

with it, this is also sometimes called the protocol stack.

Entities, Functions, Facilities and Services

These often-interchanged, somewhat vague terms refer to specific tasks or jobs performed

at various layers in the model. An N-entity is a term that refers to a specific operation or job

done at “layer N”. A function is basically the same thing. Facilities and services are what a

layer provides to the layers above it. This is often expressed in N-notation as well: the “N+1”

layer often uses a set of “N services” or “N facilities” provided by the “N” layer.

Making Sense of OSI Terminology

You may have just read all of that and said to yourself “why do they bother making this so

complicated anyway?” Well, I can't really argue with you. Remember, I did say there was

bad news here. Now that you know what all of this stuff is about, should you run into it you

won't be… uh… too badly confused. ☺

Fortunately, the use of the buzzwords above is somewhat limited. Most references are to

specific layer names or numbers, and in particular, the “N-1” and “N+1” stuff is rarely used

in discussions of real-world technologies. However, it can be very useful in explaining the

model itself, so you will see some of these terms as you read the rest of this chapter on the

OSI model.

Interfaces: Vertical (Adjacent Layer) Communication

The seven layers of the OSI Reference Model are used to split up the various functions that

are required to implement a networking system. On any given device in a network, different

software and hardware routines and devices may be functioning on any or all of these

layers simultaneously. Since in general, all of these are supposed to be working together to

implement networking functions, there is a need for layers to communicate vertically

between the layers within a particular host.

OSI Interfaces

In OSI Reference Model parlance, the mechanism for communication between adjacent

layers in the model is called an interface. Of course, the term “interface” is also used widely

in other contexts in the computer and networking worlds, since its generic meaning refers to

connecting just about anything together. However, when someone talks about an interface

between OSI model layers, that person typically refers to the process by which data is

passed between layer N of the model and layer N-1 or layer N+1.

These relationships are demonstrated in Figure 13. For example, the layer 2/3 interface is

used by a layer two and layer three protocol to pass data and control information; the layer

3/4 interface connects layers 3 and 4 together.

Note: Remember that not all layers may be implemented in every system or

protocol stack in the “real world”. So it's possible that a process that is technically

running at layer 7 might communicate with one running at layer 5. However, I am

talking about the theoretical model here.

Vertical Communication

Vertical communication is done up and down the protocol stack every time anything is sent

across the network, and of course, whenever anything is received. This occurs because the

higher levels are implemented as logical functions, in software; there is no actual physical

connection. The higher layers package data and send it down to the lower layers for it to be

sent across the network. At the very lowest level, the data is sent over the network. On the

receiving end, the process is reversed, with the data traveling back up to the higher layers

on the receiving device. The next topic dealing with horizontal communication explains

more about this logical interaction between corresponding layers.

Modularity and Inter-Layer Interactions

One of the primary goals of the OSI Reference Model is to allow the interconnection of

different implementations of various layers. Thus, the intention is to have somewhat auton-

omous individual layers that you can “mix and match”—to a point. The only way to make

this work is to have well-defined ways that the layers connect together, and that brings us

back to the matter of interfaces. Each layer must present a consistent, well-documented

interface to the layers above it so that any upper layer implementation can use the lower

layer properly.

I'll provide an example from the world of TCP/IP to illustrate what I mean. The heart of the

TCP/IP protocol suite is the Internet Protocol (IP). Whenever you use any application on the

Internet—email, Web sites, FTP, chat rooms, and so on—you are indirectly using IP.

However, you never use IP directly—you generally use one of two transport layer (layer 4)

protocols: the Transmission Control Protocol (TCP) or the User Datagram Protocol (UDP).

A standard interface exists between the network layer and the transport layer in the TCP/IP

protocol stack, which defines how IP is to be used by upper layer protocols; this enables

TCP and UDP to interface to it. Similarly, both TCP and UDP present a particular interface

to the hundreds of higher-layer protocols and applications that use them at higher layers.

Many different types of communication actually take place between layers. Control infor-

mation is passed to enable the higher layers to utilize the lower ones, and for the lower

ones to pass status and results information back to the higher ones. Data is also passed in

both directions across the interface. For transmission, it flows down to the lower layer, each

time normally resulting in data encapsulation. Upon reception, the process is reversed, with

data being sent back up across the interface from lower to higher layer.

Figure 13: OSI Reference Model Interfaces for Vertical Communication

In OSI model terminology, an interface is a conduit for communication between adjacent layers in the layer

stack.

7 Application

6 Presentation

5 Session

4 Transport

3 Network

2 Data Link

1 Physical

Layer 6/7 Interface

Layer 5/6 Interface

Layer 4/5 Interface

Layer 3/4 Interface

Layer 2/3 Interface

Layer 1/2 Interface

Key Concept: In the OSI Reference Model, an interface defines the mechanism for

vertical communication between adjacent layers. The existence of well-defined inter-

faces between layers is what permits a higher layer to use the services of any of a

number of lower layers, without requiring knowledge of how those layers are implemented.

Protocols: Horizontal (Corresponding Layer) Communication

Each layer in the OSI Reference Model has a particular role (or roles)—a set of general

tasks for which it is responsible. On each system on the network, hardware and software

are running at many of the different levels in the model. The routines doing a particular job

on machine “A” are designed to communicate with similar or complementary ones that are

running on machine “B”. This horizontal communication is the very heart of what networking

is about. It is what enables Web browsers and Web servers to talk, email applications to

exchange messages, and so much more.

OSI Protocols

Of course, all types of communication function only if everyone agrees to the same

methods of accomplishing it. Each set of rules describing one type of communication is

called a protocol. You can think of a protocol as being like a language or a set of instruc-

tions. Each function or service of a network has its own language; like human languages,

some are similar to each other while others are quite unique.

If you've done any reading at all about networks, you have probably seen the term

“protocol” many, many times. Like the word “interface”, the word “protocol” can have many

meanings. In fact, it is so fundamental to networking, and used in so many different ways,

that I have a topic devoted to it in the chapter on Networking Fundamentals.

All that aside, we must remember that the OSI Reference Model is intended to be a formal

way of describing networks. As such, the term “protocol” has a formal meaning in the

context of the model. It refers specifically to a set of communication rules, instructions and/

or procedures that describe communication between specific software or hardware

elements running at the same layer on different machines within a network.

Horizontal Communication

Let's consider how these corresponding layers communicate using protocols. First, recall

that every layer in the model, except the bottom (physical) layer, is really a program or

algorithm running on a computer. There is no way for, say, a Web browser and a Web

server to actually connect together directly—they are just software programs, after all.

Instead, the software running at various layers communicates logically. That is to say,

through the use of software and procedures, a process running at layer 5 on one machine

can accomplish logical communication with a similar process running at layer 5 on another

machine.

Since machines are only physically connected at layer 1, this means that in order for a

protocol at layer 5 to function, the data on the sending machine must “pass down” the data

through the layers between layer 5 and layer 1. The data is then transmitted over the

physical connection to layer 1 of the other machine, and “passed up” the protocol stack of

the receiving machine to layer 5. This is how the two machines are logically linked at layer

5, even though they have no physical connection at that layer.

Thus, with the exception of the actual physical connection at layer 1, all horizontal commu-

nication also requires vertical communication—down the stack on one machine, and then

back up the stack on the other. This process is illustrated in Figure 14. (The communication

doesn’t always go all the way back up the stack for each connection, however, as in the

case of routing.)

Key Concept: In the OSI Reference Model, a protocol refers specifically to a set of

rules or procedures that define communication between software or hardware

elements running at the same layer on network devices. Physical layer protocols are

responsible for the actual transmission and reception of data at layer one. Protocols at

higher layers pass data down through the layers below them to layer one for transmission,

then across the network and back up to the corresponding entity at the same layer on the

receiving device. The result is that software processes running at say, layer four on each of

two devices can communicate logically as if they were directly connected at layer four,

when they are not.

Data Encapsulation, Protocol Data Units (PDUs) and Service

Data Units (SDUs)

Protocols are what describe the rules that control horizontal communication, that is, conver-

sations between processes that run at corresponding layers within the OSI Reference

Model. At every layer (except layer one) these communications ultimately take the form of

some sort of message that is sent between corresponding software elements on two or

more devices. Since these messages are the mechanism for communicating information

between protocols, they are most generally called protocol data units (PDUs). Each PDU

has a specific format that implements the features and requirements of the protocol.

Layer Services and Data Encapsulation

As we’ve already discussed in our look at protocols, the communication between layers

higher than layer one is logical; the only hardware connection is at the physical layer. Thus,

in order for a protocol to communicate, it must pass down its PDU to the next lower layer for

transmission. We’ve also already seen that using OSI terminology, lower layers are said to

provide services to the layers immediately above them. One of the services each layer

provides is this function: to handle and manage data received from the layer above.

At any particular layer N, a PDU is a complete message that implements the protocol at that

layer. However, when this “layer N PDU” is passed down to layer N-1, it becomes the data

that the layer N-1 protocol is supposed to service. Thus, the layer N protocol data unit

(PDU) is called the layer N-1 service data unit (SDU). The job of layer N-1 is to transport

this SDU, which it does in turn by placing the layer N SDU into its own PDU format,

preceding the SDU with its own headers and appending footers as necessary. This process

is called data encapsulation, because the entire contents of the higher-layer message are

encapsulated as the data payload of the message at the lower layer.

Figure 14: OSI Reference Model Protocols: Horizontal Communication

The term “protocol” has many meanings; in the context of the OSI Reference Model, it refers specifically to

software or hardware elements that accomplish communication between corresponding layers on two or more

devices. For example, the Internet Protocol is said to be a layer 3 protocol because each device uses IP

software to communicate at layer 3. The actual transmission and reception of data only occurs at the lowest,

physical layer; higher-layer protocols communicate logically, by passing data down interfaces until it reaches

layer 1, transmitting at layer 1, and then passing the data back up to the appropriate layer at the recipient.

Physical 11 Physical

7 Application

6 Presentation

5 Session

4 Transport

3 Network

2 Data Link

Application 7

Presentation 6

Session 5

Transport 4

Network 3

Data Link 2

Layer 7 Protocol

Layer 6 Protocol

Layer 5 Protocol

Layer 4 Protocol

Layer 3 Protocol

Layer 2 Protocol

Layer 1 Protocol

Layer 6/7 Interface

Layer 5/6 Interface

Layer 4/5 Interface

Layer 3/4 Interface

Layer 2/3 Interface

Layer 1/2 Interface

Layer 6/7 Interface

Layer 5/6 Interface

Layer 4/5 Interface

Layer 3/4 Interface

Layer 2/3 Interface

Layer 1/2 Interface

What does layer N-1 do with its PDU? It of course passes it down to the next lower layer,

where it is treated as a layer N-2 SDU. Layer N-2 creates a layer N-2 PDU containing the

layer N-1 SDU and layer N-2’s headers and footers. And the so the process continues, all

the way down to the physical layer. In the theoretical model, what you end up with is a

message at layer 1 that consists of application-layer data that is encapsulated with headers

and/or footers from each of layers 7 through 2 in turn, as shown in Figure 15.

Figure 15: OSI Reference Model Data Encapsulation

Each protocol creates a protocol data unit (PDU) for transmission that includes headers required by that

protocol and data to be transmitted. This data becomes the service data unit (SDU) of the next layer below it.

This diagram shows a layer 7 PDU consisting of a layer 7 header (“L7H”) and application data. When this is

passed to layer 6, it becomes a layer 6 SDU. The layer 6 protocol prepends to it a layer 6 header (“L6H”) to

create a layer 6 PDU, which is passed to layer 5. The encapsulation process continues all the way down to

layer 2, which creates a layer 2 PDU—in this case shown with both a header and a footer—that is converted

to bits and sent at layer 1.

1 011011001001 011

L7H Data

L2H L3H L4H L5H L6H L7H Data L2F

L3H L4H L5H L6H L7H Data

L4H L5H L6H L7H Data

L5H L6H L7H Data

L6H L7H Data

Data Encapsulation in TCP/IP

The “N-1, N-2” stuff makes this seem more difficult than it really is, so let’s use a real-world

(simplified) example instead. The Transmission Control Protocol (TCP) operates at layer 4

of the OSI model. It transmits messages called segments that contain data encapsulated

from higher-layer protocols. The layer below TCP is the Internet Protocol (IP) at layer 3. It

receives data from TCP and encapsulates it for transmission.

So, in the formal language of the OSI Reference Model, TCP segments are created as layer

4 PDUs. When passed to IP, they are treated as layer 3 SDUs. The IP software packages

these SDUs into messages called IP packets or IP datagrams, which are layer 3 PDUs.

These are in turn passed down to a layer 2 protocol, say Ethernet, which treats IP

datagrams as layer 2 SDUs, and packages them into layer 2 PDUs (Ethernet frames) which

are sent on layer 1. (Actually, in some technologies further encapsulation even occurs at

layer one prior to transmission.)

On the receiving device, the process of encapsulation is reversed. The Ethernet software

inspects the layer 2 PDU (Ethernet frame) and removes from it the layer 2 SDU (IP

datagram) which it passes up to IP as a layer 3 PDU. The IP layer removes the layer 3 SDU

(TCP segment) and passes it to TCP as a layer 4 PDU. TCP in turn continues the process,

going back up the protocol layer stack. The complete process is illustrated in Figure 16.

This whole matter of passing data up and down the protocol stack, encapsulation and so on

may seem needlessly complex. It also may appear to be rather inefficient; why send a

message with so many headers and footer? However, the notion of data encapsulation is

critical to creating modular, flexible networks.

Use of PDU and SDU Terminology

The term “protocol data unit” is rather formal. You will see it used in standards and

sometimes in discussions, but more often, the “plain” message terms, such as “frame” and

“datagram”, are encountered, as discussed in the networking fundamentals topic on

messages. Similarly, data encapsulated by these messages is not normally called a

“service data unit” but rather simply the message body or payload, as discussed in the topic

on message formatting. There are cases, however, where knowing the difference between

an SDU and a PDU is important to understanding the technology. One example is the IEEE

802.11 physical layer; the 802.11 standards talk about SDUs and PDUs constantly.

Related Information: See the OSI Reference Model analogy if you want to see

an example that compares networking encapsulation to a type done in a real-

world, non-networking context.