I looked around this weekend for a simple OSI model document to share with a colleague and couldn't find a handy one that seemed simple enough. Digging around in my files, I found this one and thought I'd share it here.

Since data communications is a very complex process, it has to be broken down into manageable sections, or layers. Each layer provides for some portion of the tasks involved in the network. One of the mostly widely used approaches to this is the OSI Reference Model show in the visual. The greatest benefit to a reference model is that it provides a common language for developers to use in discussing networking. Networking is a huge and daunting task, not unlike eating an elephant. The OSI model gives us a structure so we don’t have to eat the elephant in one bite, but can break it into manageable sized pieces.

The Open Systems Interconnect (OSI) Reference Model serves several purposes:

• It provides a framework for dealing with connectivity and interoperability between dissimilar systems.
• It provides places for different protocols that perform different functions

The layers are defined in two categories. The end-to-end layers are primarily focused on the total communication from end system to end system, or user to user. The “chained layers” are present in every node of the network, whether a personal computer or a router along the path data might travel. These layers provide a chain on connections that allow data packets to be passed from node to node over a variety of disparate network architectures.




The Physical Layer
The Physical Layer describes a set of procedures or operating rules that provide for the mechanical, electrical or optical transmission of bits along a physical link. This layer defines physical connections like DTE, DCE, V.35, etc.

This transmission can be simplex (one-way), half-duplex (alternating two-way), or full duplex (two-way). It can also be a serial transmission of a single bit at a time, or parallel transmission of several bits over several parallel electrical connections. Parallel connections can be expensive, and are best used for short distances, like the ribbon cable connecting components inside a computer.

Physical layer signals are sometimes baseband, or digital signals, and sometimes broadband or analog signals.

In the previous chapter we reviewed theories by Harry Nyquist and Claude Shannon. The physical layer is where those theories are applied. Nyquist’s theorem determined that the maximum signaling rate for any bandwidth can be determined by calculating 2*Max frequency of the bandwidth. Shannon’s Law states that bandwidth and signal to noise ratio limit the information rate over a communication channel. In an electrical circuit distortion is fixed and happens to the desired signal. And example would be attenuation or the loss of amplitude over distance. Noise, on the other hand is random. It is an unwanted signal that adversely affects the transmission. In order to provide communication, we must engineer the channel to expect both distortion and noise.

The Data Link Layer
The Data Link Layer is responsible for providing error free transmission from node to node over a physical medium that is prone to errors. It only provides error detection across a single physical link. In order to accomplish this, the data is framed, or structured, and information added to allow each node to determine whether or not what was received matches what was transmitted. This is easily accomplished by performing a Cyclic Redundancy Check (CRC) against received data frames.

Basic functions performed at the Data Link layer are: (a) framing, or delineating the beginning and end of the data, (b) establishing a control mechanism for access to the media, like CSMA-CD in an Ethernet LAN, (c) detecting errors, usually by a CRC, and (d) error correction in some cases, typically in a connection oriented environment. Connectionless networks generally discard errored data.

Some examples of data link layer technologies are:

• Ethernet
• Token Ring
• Frame Relay
• ATM

The Network Layer
The Network Layers is used to transport packets across the network using some form of routing, provide for congestion control, and to establish an addressing scheme, provide either connection-oriented or connectionless services, and define an addressing scheme. In a packet environment, the network layer protocol must define the structure of a packet.

Some common network layer protocols in use today are:

• Internet Protocol (IP) in the TCP/IP protocol suite
• Internet Packet Exchange (IPX) in the Novell Netware
• Path Control layer in IBM's Systems Network Architecture

Packets are transported from one end of the network to the other by passing from node to node.

Congestion control might be handled by changing the size of buffers in various devices along the path. In many networks today, congestion is dealt with by simply discarding the packet. Higher layer protocols are robust and detect the non-delivery of packets and automatically request retransmission in many cases.

Routing protocols are the subject of numerous books, and not something we will go into here in depth. A routing protocol is just a set of rules the nodes follow. It’s a procedure used by the network to determine how to transmit a packet from node to node. They might be centralized, but a central routing database is impractical in a network like the Internet, so we use decentralized routing protocols, with some information being contained in each router.

One very crucial aspect to routing protocols today is that they must be dynamic, or sensitive to changes in the network. The Internet has many networks and hosts connected. These connections come and go as links fail, as new networks are added, and as changes occur. The network, or the nodes within the network, must learn very quickly about changes in network topology so that packets can be delivered from end to end.

The network layer protocol is where the addressing structure for a particular network is defined. IP has a specific addressing scheme, but IPX in Novell networks uses a different scheme, and DDP in the AppleTalk environment yet another. The network layer protocol defines the addressing mechanism required for each specific protocol.

Since a packet network doesn’t establish a circuit or connection, the packet itself must contain both a source and destination address. The analogy of a letter mailed at the post office is frequently used. It has both the recipient’s address for delivery, and the sender’s address for any return message, such as notification of non-delivery.

The Transport Layer
The transport layer is responsible for error control, and often flow control, on an end-to-end basis across the entire network. This is different than error control at the Data Link layer, which is only from one node to the next. This provides a mechanism for recovery from errors in the network.

Transport layer protocols that require a high degree of reliability are more complex, whereas simple protocols provide less reliable service, but with less overhead. We’ll see this comparison directly in TCP and UDP, which are both used in delivery of voice over IP packets.

The Session and Presentation Layers
The Session Layer is where management of individual user sessions, logging on and off, and recovery from some failures occurs.

The Presentation Layer deals primarily with the form and syntax of the information being transmitted and passed to the Application Layer. In human terms, this would be analogous to translating between different languages, which might use different alphabets and different phonetic intonations. Code conversion takes place in this layer.

The Application Layer
The Application Layer provides a set of protocols that deal with the meaning and form of the information. These protocols provide the syntax for a task to be completed.

Some examples of application layer protocols include

• DNS
• SMTP
• FTP

Earlier we mentioned that there are two types of software: applications and operating systems. It’s worth remembering that the operating system in a piece of hardware is an application, and most have some functions at the application layer. The operating system can also operated at other layers of the reference model depending on its construction and design.

VoIP is another application. VoIP signaling for call control and setup may operate at the application layer. Many of the QoS and security mechanisms we use to support VoIP operate at the data link and network layers, some at the transport layer.


Technorati Tags: OSI Model, network layers, OSI, IP communications