Architecture

TCP/IP is a suite of protocols used to support data communications between computer equipment on Local Area Networks (LANs) and between systems on interconnected networks. TCP/IP is also the primary protocol used on the global Internet. It is important to have a good understanding of TCP/IP to install and maintain network services on a network that is connected to the Internet, or on an Internet style Intranet. TCP/IP gets its name from two of the more important members of a large suite of protocols: Transmission Control Protocol (TCP) and Internet Protocol (IP). As these names suggest, the primary purpose of TCP/IP is to provide communications across interconnected networks. The TCP/IP protocol suite is popular because it has the features used to meet worldwide networking needs. These features include: Hardware and software independence Publicly documented, freely available standards A large and flexible addressing scheme The standards for TCP/IP protocols are developed and maintained independently from any specific hardware type or operating system, and have been implemented for most popular platforms. TCP/IP is available on Intel and Macintosh personal computers, DEC servers, HP and Tandem minicomputers, IBM mainframes, and most other major computer systems. TCP/IP is the native network protocol for UNIX and is available for all Microsoft operating systems, the Macintosh O/S, IBM OS/2, and most popular operating systems. It can be run on many different network types, including Ethernet, token ring, X.25, serial lines, and wireless systems. The documentation on TCP/IP protocol standards is freely available and widely published. Most of the information on these standards is published as Requests For Comments (RFCs) which are available at the Internet Network Information Center (InterNIC) as well as many Universities and other networking information centers. Some of the RFCs that are referenced in this document are included in the section at the end of this technical reference for the reader's convenience. The addressing scheme implemented in IP version 4 (IPv4), which is used throughout the Internet, allows for over a billion devices to be uniquely identified and contacted across a network. The design allows for networks as simple as two computers connected by a single cable, to networks as complex as the global Internet. It is traditional when teaching TCP/IP to begin by discussing the OSI Reference Model of networking, to compare this model to the TCP/IP networking architecture model, and to discuss all protocols as they relate to the various layers of the model. In actual practice the OSI Reference Model is seldom used when working with networks, and most of the protocols in the TCP/IP suite cross the traditional layer boundaries. Partly in deference to tradition, and partly just to lend some degree of organization to this document, the authors have chosen to organize the sections along the TCP/IP networking model, and have included the obligatory discussion of OSI and the comparison to the TCP/IP model. To discuss network communication it is often necessary to use terminology that is unique to the networking world. A model developed by the International Standards Organization (ISO) is normally used when discussing data communications protocols. It is called the Open Systems Interconnect (OSI) Reference Model. This model defines seven distinct layers, each with its own specific functions. TCP/IP architecture is based on the OSI model, but as is often the case when applying theory to a practical application, TCP/IP does not follow the OSI model exactly. TCP/IP is usually described as having fewer layers than the OSI model, and many of the functions that are distinct to a layer in OSI cross layer boundaries in TCP/IP.