4-1 Computer networks

In the nineteen-sixties, almost immediately after computers became widespread, the desire grew to connect computers together. Originally the aim was to allow the exchange of information (programs, data files) between the users of different mainframes. When smaller computers became available and spread in larger numbers within a single organization, connections between these small computers became attractive for the shared use of resources (processing time) as well. The resulting systems, in which computers of possibly different type and size are connected but fully retain their autonomy, are called computer networks. Computer networks, world-wide networks of (large) computers are referred to as wide area networks, and campus-wide networks of (small) computers are referred to as local area networks.

The complete design of a computer network and the networking software requires the solution of many problems and is a very complicated matter. To reduce the design complexity a network architecture is viewed as a hierarchy of layers. Each layer (or level) consists of protocols that use functions of the lower layers and provide services to the higher layers. Each layer performs well-defined functions, and can be of limited complexity. Together the layers implement the facilities of the network.

The approach to network architecture and implementation was standardized in the International Standards Organization's OSI [1] reference model for Open Systems Interconnection. The model contains seven layers and was designed so as to prevent the networking software of different vendors being largely incompatible.[2] The following layers are distinguished. The physical layer provides protocols for transmitting bits via a line. The data link layer implements message transfers and handles transmission errors. In the network layer configuration-dependent functions are implemented, including, e.g., the routing of messages. The routing problem requires that a path is found to transport in formation between sites that are not connected directly by a channel. As the network layer may route subsequent messages via different paths, messages may arrive in an order different from the order in which they were sent. Also, due to the possibility of intermediate computers in the network going down, messages may be lost or duplicated (as the result of a retransmission). A next layer in the hierarchy, the transport layer, serves to deal with these problems. The session layer concerns the problem of reliably opening and closing connections between processes in the network. The presentation layer performs the necessary data compression and conversion. The application layer contains protocols for the exchange of information between distributed user programs (applications).

In a hierarchical architecture the model of distributed systems can be used for the design of protocols in each of the levels. Only the assumptions about the system and the problems to be solved differ from level to level. For example, in the data link layer the possibility of transmission errors is considered, but in the higher level protocols this possibility is ignored, in the network layer it is assumed that the processes are not all directly connected by a channel (otherwise the routing problem would be trivial). In higher levels, however, the assumption can be made, because message passing between every pairs of processes is implemented by the network layer. In the transport layer only the two endpoints of a channel are considered; the distributed system it applies to consist of two processes only.

WORDS AND PHRASES

channel      通道

hierarchy      层次，等级结构

routing problem路由问题

vendor         厂商，自动售货机

NOTES

[1] International Standards Organization—ISO 国际标准化组织

[2] The model contains seven layers and was designed so as to prevent the networking software of different vendors being largely incompatible.

这种模式包括七层，之所以如此设计是防止不同厂商生产的网络软件互不兼容

词汇

课文