The TCP/IP and OSI Models
TCP/IP and OSI are two well-known sets of communication protocols that have similarities and differences. This white paper discusses the history of each model, along with the characteristics and structures that make each one unique.
Two well-known protocol suites began working toward their standardization processes at about the same time. One was being developed under contracts to the Defense Advanced Research Projects Agency (DARPA). It eventually became TCP/IP, named for its core protocols: Transmission Control Protocol (TCP) and Internet Protocol (IP). Its cousin grew up under the watchful eyes of the International Organization for Standardization (ISO)1. Its name was more focused on its function: Open Systems Interconnection (OSI)1. OSI is composed of two parts. Part one, called the Basic Reference Model, is also known as the 7-layer model, or the OSI model. The other part is a set or suite of specific protocols for each layer.
In 1981, the Network Information Center (NIC) identified TCP/IP as its Internet standard in Request for Comments (RFC) 791 and 793. As of August, 1983, Department of Defense (DoD) set it as Military Specification 1778. In 1984, ISO specified OSI as ISO standard 7498. At the same time, the Consultative Committee on International Telephone and Telegraphy (CCITT), now known as the International Telecommunications Union Telecommunication Standardization Sector (ITU-T), designated the OSI Model as standard X.200. In 1987, after a subset of the OSI suite known as the Government Open Systems Interconnection Profile (GOSIP) became a Federal Information Processing Standard (FIPS), DoD sent out a letter declaring GOSIP "an experimental co-standard to the DoD protocols" as a result of the standards work done by the American National Standards Institute (ANSI) and the National Bureau of Standards (NBS). Though TCP/IP was and is in the public domain, and so free to use, GOSIP held the potential for large government contracts. This led to many companies wasting vast sums of money trying to produce OSI software while the networking community moved ahead with developing TCP/IP protocols, and then writing the standards. By 1994, after repeated attempts to make the GOSIP solution function, the OSI FIPS was finally rescinded.
Both protocol suites (TCP/IP and OSI) use a layered approach.
They are both designed to provide end-to-end communication.
Both have similar Transport and Network layers.
The suites both assume packets will be switched, meaning the packets can take different paths to their destination.
Both solutions require that the network professionals using them must understand them in detail.
With the common use of the OSI model as a way to describe the layers and their functions, network professionals must know both models.
The application layer in TCP/IP handles the responsibilities of multiple layers in the OSI model.
The OSI model numbers and names its layers, whereas the TCP/IP stack only names the layers.
Unlike the transport layer in OSI, TCP/IP only guarantees reliable delivery of packets when TCP is the chosen protocol.
OSI has much more complexity in its 7 layers than TCP/IP has in its 4 layers.
In TCP/IP, protocols are deliberately designed to have more layer flexibility than the strict layers of the OSI model.
TCP/IP functions are implemented, then standardized. OSI is standardized in concept only, though some functions work.
OSI has more limited Network Management and Network Security.
TCP/IP Stack Layers
The four layers in TCP/IP are, from the top down: are as follows.
The Application Layer has the responsibility for authentication, data compression, and end-user services such as terminal emulation, file transfer, e-mail, web browsing/serving, and other network control and management services. An application header and following data are packaged as a message.
The Transport Layer, sometimes also called the Host-to-Host Layer, handles end-to-end communications, error handling, flow control, and connection-oriented or connectionless communication. An application message and a connection-oriented TCP header combine into a TCP segment. When a connectionless UDP header is attached to an application message, the result is a UDP datagram.
The Internet Layer, also known as the Internetwork Layer, does the heavy lifting for the network by supporting logical addressing, mapping logical addresses to physical addresses, routing decisions, subnetting, multicasting, logical error reporting, and network diagnostics. Adding an IP header to a TCP segment or a UDP datagram creates an IP datagram.
The Link Layer, which is also called the Network Interface Layer, Network Access Layer, or Local Link, interfaces the rest of the protocol stack with the link protocols and the local communications hardware. It functions locally and may be different along each network from one host or system to another. Unlike the other three layers, other standards organizations, like the Institute for Electrical and Electronics Engineers (IEEE), set the standards for this layer, for example, for Ethernet and Wi-Fi. Adding a link layer header to the front of an IP datagram and a checksum or Frame Check Sequence (FCS) to the end frames the data and is called a frame. Some other link layer protocols identify the created package by adding their header to the IP datagram as a packet. The generic name is also "packet," which is left from the early days of data communication.