Second, computer network architecture

first, the formation of computer network architecture

1, why layering?

when dealing with complex systems:

A clear structure allows us to define a specific part of a large complex system and explore its various parts of the relationship
Modularity makes system maintenance and upgrades simple

Change the specific implementation of one layer of service transparent to the rest of the system (no impact)
e.g. changing the boarding gate procedure (does not affect the rest of the system)

the drawbacks of layering:

Layers may repeat lower levels of functionality ... 2, the question is raised

Network system is very complex, contains a large number of hardware and software, and the communication of computer systems must be highly coordinated to work properly, "layered" can transform the problem into a number of small and easy to deal with local problems. 3, an example of the concept of dividing hierarchy

Computer 1 sends a file over the network to computer 2.

The work to be done will be divided into the following. 1) Two computers take the file transfer module as the highest tier, and the rest is responsible for the following modules. Its work is directly related to the transfer of documents. :

Be sure that the other party is ready to receive and store the files.

Both parties coordinate a consistent file format.

2) Set up a Communication service module to ensure that files and file transfer commands are reliably exchanged between the two systems. That is, the file transfer module on the upper level leverages the services provided by the following communication services Module.

3) construct a network access module, which is responsible for doing the work related to the network interface details (for example, specify the frame format of the transmission, the maximum length of the frame, etc.), and provide services to the upper layer, so that the above Communication Service module can complete the task of reliable communication.

4. Benefits of Layering

(1) Each layer is independent.
(2) Good flexibility.
(3) The structure can be divided open.
(4) Easy to implement and maintain.
(5) Can promote standardization work.

Note: Number of layers to be appropriate

(1) Too few layers will make each layer of the protocol too complex.
(2) The number of layers is too many, it will be more difficult to describe and synthesize the functions of each layer of systems engineering tasks. 5, the formation of computer network system structure

In 1974, IBM in the United States announced the SNA for the System network architecture it developed
(System Network Architecture). It is one of the most widely used network architecture in the world. Soon some big companies are rolling out their own architectures. 1) Two types of standards

The International Organization for Standardization (ISO) set up a specialized agency research network architecture in 1977. A standard framework, the basic Reference Model for open Systems Interconnection, is proposed, which attempts to interconnect various computers worldwide OSI/RM
(Open Systems interconnection Reference Model), referred to as OSI, formed the ISO7489 Standard, the seven-tier protocol architecture, in 1983.

about the Open Systems Interconnect Reference Model OSI/RM:

As long as the OSI standard is followed, a system can communicate with any other system that is located anywhere in the world that also follows this same standard.

The OSI has failed in terms of marketability, for the following reasons: OSI experts did not have a commercial drive to complete the OSI standard, the OSI protocol was overly complex and inefficient to operate, and the OSI standard was too long-established to allow equipment manufactured under the OSI standard to enter the market in a timely manner; the OSI layer Sub-division is not reasonable, and some functions are repeated in multiple levels.

The legal international standard OSI has not been recognized by the market. Non-standard TCP/IP is now the most widely used. TCP/IP is often referred to as de facto international standards.

The OSI concept is clear, the system price is low, the theory is more complete, but the realization is complex. The TCP/IP protocol now does not actually have a complete architecture. 2) Protocol

A protocol is a rule, standard, or convention established for the exchange of data in a network. A network protocol consists mainly of the following three elements:

(1) syntax, i.e. the structure or format of data and control information;
(2) Semantics, that is, what kind of control information needs to be emitted, what action to complete and what response to make;
(3) synchronization, which is a detailed description of the event implementation sequence.

Two forms of existence of the protocol: literal description, program code 3) computer network architecture Concept

The architecture of the computer Network (architecture) is a collection of the various layers of the computer network and their protocols.

Architecture is the precise definition of the function that the computer network and its components should perform.

Implementation (Implementation) is the question of which hardware or software is used to accomplish these functions, following this architecture.

The architecture is abstract, and the implementation is specific and is really running on the computer hardware and software. Two, computer network architecture with five-layer protocol

The OSI architecture is both complex and impractical, but the concept is clear and the theory is more complete. The TCP/IP protocol has been recognized worldwide, but its architecture is incomplete. Therefore, in the study of the principle of computer networks often take a compromise approach, combining the advantages of the two, using a five-layer principle system structure.

Host 1 sends data process analysis to host 2:

third, entities, protocols, services, and service access points 1. Basic Concepts 1) Entity: represents any hardware or software process that can send or receive information.

2) Protocol: is a collection of rules that control the communication of two peer entities. 3) Service: Under the control of the Protocol, the communication between the two peers enables the layer to provide services up to one level. To implement this layer of protocol, you also need to use the services provided in the following layer.

4) Data Unit

A, Protocol data unit PDU

A unit of information that is exchanged between layers of peer entities in different endpoints for the implementation of this layer protocol.
PCI: Protocol control information UD: User data

B, Service data unit SDU

The (N) entity is the required data unit for the function requested by the completion (N+1) entity.

5) Service access point SAP: The place in the same system where two adjacent layers of entities interact (that is, exchanging information), often referred to as (service access points).

6) Service primitives: Upper layer the services provided by the lower layer must be exchanged with the lower layer commands, which are called Service primitives in the OSI.

Four types of service primitives:
(N)-primitive NAME request (A0,..., an-1)
(N)-primitive NAME Indication (A0,..., an-1)
(N)-primitive NAME response (A0,..., an-1)
(N)-primitive NAME confirm (A0,..., an-1) 2, difference between service and agreement

Protocols and services are conceptually different.

First, the implementation of the Protocol guarantees the ability to serve up a layer. Service users at this level can only see the service and cannot see the following protocol. The following protocols are transparent to the service users above.

Second, the protocol is "horizontal", i.e. the protocol is the rule that controls the communication between peer entities. But the service is "vertical", that is, the service is provided from the lower level to the upper layer through the inter-tier interface.

The protocol is complex:

The agreement must anticipate all unfavourable conditions beforehand and not assume that everything is ideal and smooth.

It is important to check carefully whether the protocol is designed to handle all adverse situations.

It should be noted that there are, in fact, very individual adverse circumstances that were not foreseen in the design of the agreement. When this happens, the protocol fails. Thus, in practice, agreements often only deal with the overwhelming majority of adverse situations.

Well -known protocol examples

The blue Army, which occupies two summits, fights with the White Army stationed in the valley. The strength of the comparison is: a peak of the Blues can not beat the White Army, but the two peak of the Blue Army to combat the case. A blues on the top of the Hill intends to launch an attack on case the next day. The message is then sent to a friendly on the other hill. But the communication line is very bad, the message error is very likely. Therefore, a friendly must send a confirmation message to the incoming message. However, confirmation of the message may also be wrong. Could it be possible to devise a protocol that would allow the Blues to achieve a concerted battle and thus a certain (i.e. 100) victory.

Such a protocol cannot be implemented because the protocol that returns a message after it receives the message is infinitely looped. 3. Connection-oriented services and non-connected services

From the point of view of communication, the services provided by each layer can be divided into two categories, namely, connection-oriented and non-connected. 1) connection-oriented services (connection-oriented)

The virtual circuit is serviced in the network layer.
The transport layer is a TCP service.
Working features with Phone 2) No connection service (connectionless)

The data is "best-effort delivery".
Working features with regular mail four, TCP/IP architecture 1,tcp/ip four-layer model

2,osi vs TCP/IP

OSI vs. TCP/IP architecture comparison:

The OSI concept is clear, the system price is low, the theory is more complete, but the realization is complex.

TCP/IP is very different from OSI in the handling of some problems: TCP/IP begins with the interconnection of multiple heterogeneous networks in mind. TCP/IP begins with both a connection-oriented service and a non-connected service. TCP/IP has better network management functions.

- of course, TCP/IP also has shortcomings: The concept is not clear that the common poor network interface layer has no substantive content

client processes and server processes communicate using the TCP/IP protocol:

a more powerful computer runs multiple server processes at the same time: