Charles M. Kozierok The TCP-IP Guide

Подождите немного. Документ загружается.

larger ones. The best example of this is the upcoming change to the Internet Protocol.

While DNS is important for the 32-bit addresses used in IP version 4, it's even more

important for dealing with the enormous 128-bit addresses of IP version 6.

Figure 230: Internetwork Access With A Name System

When an internetwork is equipped with a name system, the user no longer needs to know the address of a

device to access it. He or she enters the name and the name system converts it into an address automatically,

like a computerized “rolodex”, as I have shown here. ☺ The name system then passes the address to the

client software which uses that address to access the requested resource as if the user had entered it directly.

Ente r Name

Human User Client Server

Request Resource

Process Request

Generate Reply

Process Reply

Access Resource

Name System

Process Name

Request

Resolve Name To

Address

Key Concept: Networking name systems are important because they allow devices

to be assigned efficient numeric addresses, while still enabling humans to access

them using names that are easier to remember. Name systems become more

important as you increase the size of the network, the address or the user base. They are

also more essential when the user base is limited in skill or experience.

Name System Functions: Name Space, Name Registration and

Name Resolution

While the difference between numeric addresses and symbolic names is very significant to

the users of network devices, it's important to remember that both numbers and names

really serve the same basic purpose: device identification. Even when we use a name

system to make devices easier to access, the computers themselves will still normally need

to use the underlying numeric identifier. In essence, then, every device will end up with (at

least) two identifiers: a number and a name.

The fact that devices end up with multiple identifiers is what allows both people and their

machines to use the method of identification they prefer. However, it means that there must

be ways of managing the assignment of names to devices, and of converting between

them. A name system therefore involves more than just slapping names on computers. It

must be, in fact, a complete system that allows names to be used by the humans while

numbers continue to be used by the devices.

Overview of Basic Name System Functions

At the highest level, I consider a name system's jobs as encompassing these three basic

functions:

☯ Name Space: The name system defines a name space for the networking system

upon which it runs. The name space, also sometimes called a name architecture,

describes the rules for how names are structured and used. It also defines how the

name of one device is related to the names of other devices in the system, and how to

ensure that no invalid names are given that would cause problems with the system as

a whole.

☯ Name Registration: To implement the name system, a name must be assigned to

each device on the network. Like any addressing system, a name system cannot work

properly unless every name on the system is unique; we need some way of managing

how the names are assigned so the result is sensible. The process of linking specific

names to particular devices is usually called name registration.

☯ Name Resolution: As I mentioned above, even though humans like symbolic names,

computers usually have little use for them. It is necessary to define a mechanism by

which a device's symbolic name can be translated into its numeric address. This

process is usually called name resolution.

Relationships Between Name System Functions

The name space is more of a “descriptive” function; it is a definition of how names work in

the system. Name registration and resolution are more “active” functions, with each name

system including one or more specific procedures for how these jobs are carried out. Name

registration and resolution are in some ways complements of each other, so certain regis-

tration techniques are most often associated with particular resolution methods. In turn, the

types of registration and resolution methods that are possible depends on the name space

and in particular its architecture. These relationships are shown in simplified form in Figure

231.

The processes of registration and resolution can be either quite plain or fairly complicated,

depending on the type of name system used. Simple name systems are largely manual in

operation, easy to understand and best used in smaller networks. Larger, more complex

networks and internetworks require more sophisticated methods of registration and

resolution, which work with less need for administrator intervention and scale better as new

machines are added to the network.

Figure 231: Name System Functions

This diagram shows the relationships between the three main functions of a name system. The name space

defines the structure of the name system and the rules for creating names. The name space is used as the

basis for the name registration method, which defines the mappings between names and addresses.When a

user wants to access a device by name, a name resolution method is used to consult the name space,

determine what address is associated with a name, and then convert the name to an address.

Name Space

Name Registration Name Resolution

Requests to Associate

Device Names

To Addresses

Name/Address

Mappings

Name Resolution

Requests and Replies

Name/Address

Mappings

I should also point out that while I consider name registration and name resolution as

“functions” at the highest level, they are probably better thought of as sets of functions.

Name registration is necessarily tied into issues such as name system administration and

management, and understanding resolution requires that we look at a number of important

implementation issues in the areas of efficiency and reliability. The next few topics expand

on this overview by considering each of these three functions in more detail.

Key Concept: A name system consists of three theoretical high-level functions: the

name space, which describes how names are created and organized; the name

registration technique, which is used to set up relationships between names and

addresses; and the name resolution method, which is responsible for translating names to

addresses.

Name Spaces and Name Architectures (Flat and Hierarchical)

The main idea of a name system is to provide a way to identify devices using symbolic

names. Like any identification mechanism, before we can use the system we must define

the way that identification will be performed. Numeric addressing schemes (like IP

addresses) have rules for how addresses are created, and assign addresses to each

device from their address space. In a similar way, devices in a name system are given

names from the system's name space.

Name Space Functions

Of the three main functional components of a name system, the name space is the most

abstract. It is also the most fundamental part of the system, since it actually describes how

the names are created. There are several aspects to what the name space defines in a

name system:

☯ Name Size and Maximum Number of Names: The name space specifies the number

of characters (symbols) that comprise names. In so doing, it also defines the maximum

number of names that can appear in the system.

☯ Name Rules and Syntax: The name space specifies what characters or symbols are

allowed in a name. This is used to allow legal names to be chosen for all devices while

avoiding illegal names.

☯ Name Architecture and Semantics: Each name space uses a specific architecture

or structure, which describes how names are constructed and interpreted.

Name Architectures

The concepts of name size and name syntax are relatively straight-forward. The last item,

name architecture, is probably the most important differentiating characteristic of name

systems. For this reason, name spaces are sometimes even called name architectures.

The architecture of the name space determines whether names are assigned and used as

a simple unstructured set of symbols, or whether they have a more complex internal

structure. In the latter case, the name space also must define how elements of a particular

name are related to each other.

Theoretically, many different name architectures are possible. In practice, most fall into one

of two categories: flat and hierarchical architectures.

Flat Name Architecture (Flat Name Space)

Names are assigned as a sequence of symbols that are interpreted as a single, whole label

without any internal structure. There is no clear relationship between any name and any

other name.

An example of this sort of architecture would be a name system where computers are given

unstructured names like “Engineering Workstation 1” or “Joanne's PC”, as shown in the

example in Figure 232.

Hierarchical Name Architecture (Hierarchical or Structured Name Space)

In this architecture, while names consist of a sequence of symbols, these symbols are

assigned using a specific and clear structure. The name consists of discrete elements that

are related to each other usually using hierarchical “parent/child” semantics. There are

many naming architectures in various contexts that use this type of hierarchical structure;

for example, consider how a large company might set up an organization chart and name

the executives and officers in the organization. One hypothetical example of a hierarchical

name architecture is illustrated in Figure 233.

The best-known real world example of a hierarchical name space is the name space of the

TCP/IP Domain Name System, which uses text labels separated by periods (“dots”) to form

an internal structure. All the names in the system are organized into a structure, and a

Figure 232: Flat Name Architecture (Flat Name Space)

This diagram shows an example of a flat name architecture. There is no structure that organizes the names or

dictates how they must be constructed. Logically, each device is a peer of each of the others.

XYZ Industries

Jo In Sales

Jose's PC

Jaclyn

Jean, France

Jalil's Machine

Julie J.

John's Notebook

Jill

EU-Service-Jorge

Jae's PC

particular device's place in the structure can be determined by looking at its name. For

example, “www.tcpipguide.com” refers to the World Wide Web server for The TCP/IP

Guide, which is named under the umbrella of “COMmercial” companies.

Comparing Name Architectures

As we will see in the next two topics, the architecture of the name space is intimately related

to how names are registered and managed, and ultimately, how they are resolved as well. A

flat name space requires a central authority of some sort to assign names to all devices in

the system to ensure uniqueness. A hierarchical name architecture is ideally suited to a

more distributed registration scheme that allows many authorities to share in the regis-

tration and administrative process.

Figure 233: Hierarchical Name Architecture (Structured Name Space)

This diagram contains the same devices as Figure 232, but they have been arranged using a hierarchical,

structured name architecture. In this case, the organization has chosen to structure its device names first by

facility location, and then by department. Each name starts with something like “USA-Service-” or “EU-Mfg-”.

This provides immediate benefits by providing local control over device naming without risk of conflicts. If

someone named John were hired into the USA sales force, his machine could be named “US-Sales-John”

without conflicting with the machine owned by John of the European sales force (“EU-Sales-John”.) The

structure also makes it easier to know immediately where a device can be found within the organization.

XYZ Industries

United States

Operation

European Union

Operations

USA-ServiceUSA-Sales EU-Mfg EU-Sale s EU-Service

USA-Sales-Jo

USA-Sales-Jose

USA-Service-Jaclyn

EU-Mfg-Jean

EU-Mfg-Jalil

EU-Sales-Julie

EU-Sales-John

EU-Sales-Jill

EU-Service-Jorge

EU-Service-Jae

All of this means that the advantages and disadvantages of each of these architectures are

not a great mystery. Flat name spaces have the advantage of simplicity and the ability to

create short and easily remembered names, as shown in Figure 232. However, they do not

scale well to name systems containing hundreds or thousands of machines, due to the diffi-

culties in ensuring each name is unique. For example, what happens if there are four

people named John who all try to name their computers “John’s PC”? Another issue is the

overhead needed to centrally manage these names.

In contrast, hierarchical name spaces are more sophisticated and flexible, because they

allow names to be assigned using a logical structure. We can name our machines using a

hierarchy that reflects our organization’s structure, for example, and give authority to

different parts of the organization to manage parts of the name space. As long as each

department is named uniquely and that unique department name is part of each machine

name, we don’t need to worry about each assigned name being unique across the entire

organization, only within the department. Thus we can have four different machines named

with their department name and “John”, as Figure 233 demonstrates. The price of this flexi-

bility is the need for longer names and more complexity in name registration and resolution.

Key Concept: The two most common types of name architecture are the flat name

space and the hierarchical name space. Names in a flat name space are all peers

with no relationship; in a hierarchical architecture, a multi-level structure is used to

organize names in a specific way. The flat system is simpler and satisfactory for small

networks, while the hierarchical name space is more flexible and powerful, and better-

suited to larger networks and internetworks.

Name Registration Methods, Administration and Authorities

It seems obvious that for our name system to be implemented, we need some method of

assigning names to each of the devices that will use the system. Just as a name system

has a name space that is comparable to an addressing system's address space, it also

must implement a set of rules and procedures for assigning names, just as an addressing

system assigns addresses. This is called name registration.

Name Registration Functions

In general, name registration encompasses the following concepts and tasks:

☯ Name Assignment and Guaranteeing Uniqueness: The core task of the name

registration process is of course assigning names to devices. Like all identification

schemes, a key requirement of name registration is ensuring that each name is

unique. Duplicated names cause ambiguity and make consistent name resolution

impossible.

☯ Central Registration Authority Designation: Ensuring uniqueness of names

requires that there be someone “in charge” of the name assignment process. This

central registration authority may be a single individual that maintains a file containing

names, or an organization that is responsible for the overall name registration process.

The authority is also charged with resolving problems and conflicts that may arise in

registrations.

☯ Registration Authority Delegation: In smaller name systems, the central registration

authority may be responsible for the actual registration process for all devices. In

larger, hierarchical name systems, having this process centralized is impractical.

Instead, the central registration authority will divide the name space and delegate

authority for registering names in different parts of it to subordinate organizations. This

requires a delegation policy to be developed and implemented.

☯ Defining Hierarchical Structure: When a hierarchical name space is used, the

central authority is responsible for defining how the structure will look. This in turn

dictates how names can be registered in different parts of the hierarchy, and of course,

also impacts how authority is delegated.

Impact of Name Space Architecture on Name Registration

The complexity of the name registration process depends to a great extent on the size and

complexity of the name system as a whole, and in particular the architecture of the name

space. In a simple name system using a flat name space, registration is usually accom-

plished using a single authority. There is no structure, of course, and usually no delegation

of authority, so there isn't much to registration.

For hierarchical name systems, name registration is tied tightly to the hierarchy used for

names. The central authority defines the structure of the hierarchy, and decides how the

hierarchy is to be partitioned into subsets that can be independently administered by other

authorities. Those authorities may in turn delegate subsets of their name spaces as well,

creating a flexible and extensible system.

This ability to delegate authority for name registration is one of the most powerful benefits of

a hierarchical name space. For example, in the TCP/IP Domain Name System, a central

authority is responsible for name registration as a whole; it is in charge of deciding what

top-level domains, such as “.com”, “.edu”, “.info” and “.uk” are allowed to exist. Authority for

managing each of these subsets of the worldwide hierarchy is then delegated to other

organizations. These organizations continue the process of dividing the hierarchy as they

see fit. Eventually each organization is able to decide how it will name its own internal

systems independently; for example, IBM can register names in any way it sees fit within

the “ibm.com” name.

Name Registration Methods

There are several common methods by which the actual process of registration is carried

out. Each of these has its strengths and weaknesses, and again, some are better-suited to

flat name spaces and some to hierarchical ones.

Table Name Registration

In this technique, name assignments are maintained in a table by an administrator. When

names need to be added, deleted or changed the table is edited.

This technique is usually associated with small, flat name space name systems, and has

the same benefits and drawbacks as flat architecture in general: it is simple and easy to

implement, but doesn't scale well to larger systems. With a dozen machines, having

someone edit name registration tables is practical; with thousands of machines it is not. It is

also not conducive to a hierarchical system where there are multiple authorities, because

the table needs to be kept in one place.

In larger internetworks, tables may be used as an adjunct to one of the more sophisticated

techniques below.

Broadcast Name Registration

This is a “trial and error” technique; a device that wants to use a particular name sends out

a message to all other devices on the network asking to see if anyone else is already using

it. If so, it chooses a different name. If not, the name is considered registered and can then

be used.

This technique is more sophisticated than using tables, but is still limited to use in relatively

small systems. It is not practical to attempt to broadcast to thousands of systems, and this

method could not be used over the Internet, since there is no way to broadcast to every

device on an internetwork.

Database Registration

A database of name assignments is maintained. To register a name, a request must be

made to have the name assignment added to the database. If the authority for the name

system is entirely centralized, the database will be centralized and maintained by that

authority. If authority for parts of the hierarchy is delegated, then a distributed database is

used for registration, with each authority maintaining the part of the database describing

their section of the hierarchy.

This is the most sophisticated technique, and one normally associated with hierarchical

name systems like DNS. It has several benefits, including flexibility, reliability and distri-

bution of maintenance effort; its main drawback is complexity.

Key Concept: Name registration is the process by which names are linked to

addresses in a name system. It encompasses activities such as central registry

authority designation and delegation, and name space structure management. The

most common methods of name registration, in order of both increasing capability and

complexity, are manual table maintenance, broadcast registration and database

registration.

We have just seen that the choice of name space and architecture immediately has an

important impact on how name registration is done. The next topic will show how name

resolution is closely related to name registration and architecture as well.

Name Resolution Techniques and Functional Elements of A Name

Resolution System

As we discussed earlier in this section, using a name system creates two “parallel” identifi-

cation systems for computers: the numbers used by machines and the names used by

people. The job of the name system is to integrate these two schemes. Name registration

allows humans to specify which machines use which names. This is only half the process,

however; we also need a way for machines to take a name given to them by a human and

translate it into the “real” address it can actually use for communication. This is called name

resolution.

Name resolution, also sometimes called name translation, mapping or binding, is the most

well-known aspect of name systems, because it is where most of the “heavy lifting” of a

name system occurs. The name space is generally set up once, and name registration

occurs infrequently: only when names must be created or changed. In contrast, every user

of a name system instructs the machines he or she uses to perform name resolution,

hundreds or even thousands of times a day.

Name Resolution Methods

There are several different techniques that can be use for name resolution. How this

function is implemented depends a great deal on the other two name system functions,

name space and name registration. As you might imagine, a simple name system with a

simple name registration method will most often use a simple resolution method as well.

Complex hierarchical systems with distributed databases require more sophistication in

how names are resolved.

There are three common name resolution methods.

Table Name Resolution

The table used for table-based name registration is consulted by a device when resolution

needs to be performed. The table tells the device how to transform the name of the

machine it needs to contact into an address.

This technique obviously corresponds to table name registration; it is the simplest and least

capable of the three methods. Table name resolution is suitable for standalone use only in

very small name systems, but can be a supplement to other methods as well.

Broadcast Name Resolution

When a device needs to resolve a name, it broadcasts a query that says something to this

effect: “I need to send to the device named 'X', who is that?” The device whose name is 'X'

responds back saying “I'm 'X' and my numeric address is 'N'”.