OSI Model

Open Systems Interconnection Model

The model is a theoretical object, most often followed loosely and not to the letter, that breaks down the functions of a network into seven layers.

Protocol - Defines almost every aspect of the language that is used for computers to communicate.

De jure standard - protocol designed by one company or organization and is maintained by that company.  De jure protocols can either be open or closed standards.  An open standard is one in which the specifications of the protocol are released.

De facto - protocol controlled by the entire industry and this is known as industry standard.  All de facto protocols are open standards.

Protocol Stacks - The layered functionality of the different protocols in the OSI model is called a protocol stack.  When data is sent from a source device down the OSI model, each layer attaches its own header to that information.

 

OSI Layer Function Data Name
Application Interface between the user's application and the network Messages
Presentation Negotiates data exchange formats Packets
Session Allows users to establish connections using easily remembered names Packets
Transport Provides end-to-end, reliable connections Datagrams and segments
Network Routes data through a large internetwork Datagrams
Data Link Determines access to the network media Frames
Physical Transforms data into bits that are sent across the physical media Bits

Application Layer - This layer is the interface between the user's application and the network.  This should not be confused with the actual application that the user is running.  This layer enables the user to interact with network.

 

Presentation Layer - Establishes the format for data exchange.  Also handles character set and numeric translations and data compression.

 

Session Layer - This layer lets users establish a connection - called a session - between devices and manages dialogue.  Session can be set up as follows:

Half-duplex:  One device talks while the other listens.  Like a CB radio.
Advantages:  Costs less and enables two way communication.

Disadvantages:  Costs more than simplex and only one device communicate at a time.

Simplex:  One-way communication only.
Advantages:  Cheapest method

Disadvantages:  One way only.

Full-duplex:  Two-way communication.  Phone is an example of duplex.
Advantages:  Two-way communication.
Disadvantages:  Most expensive because two bandwidth channels necessary.

 

Transport Layer - This layer carries on a dialogue with the destination device by using message headers and control messages to ensure all the packets are received.  It takes information from the session layer and breaks it up if necessary and when receiving data it puts it back together.  It ensures that when data is broken up that is put back together in the proper sequence.

 

Network Layer - This layer routes information from sender to receiver, converts data into packets and uses connectionless transmissions (meaning it is not this layer's job to be sure things get there or not).  The network layer uses the following to determine the best way to send a packet:

Switching:  Circuit switching is where a dedicated connection is made between two communicating devices.  Advantages - no congestion and delay.  Disadvantages - inefficient use of media and long wait for connection.  Message switching is where data is moved from device to device in its entirety.  Advantages - media used more effectively and congestion can be controlled and prioritized.  Packet switching is where data is broken into small pieces and routed from device to device.  Datagram packet switching tags information with address so no dedicated connection is needed.  Virtual circuits are similar to actual except the media is shared.

Routing:  For packets to be routed across networks there needs to be a table set up to show the shortest route between two networks.  These tables can be created and maintained either dynamically or statically.  Static tables are usually setup by an administrator.  Dynamic tables are defined by one of two methods:  distance vector or link state.  Distance vector simply calculates the number of hops between destinations.  Link state takes into account hop count, link speed, latency and congestion.

Addressing:  A device on a network has not only a device address but a network address.  This network address lets the layer know if the device is on the same network as the destination device.

 

Data Link Layer - The main purpose of this layer is to provide a reliable method of transmitting data across the physical media.  This layer packages data into frames, transmits data sequentially and processes acknowledged frames sent from receiver.  This layer processes individual packets where the transport layer assures total data reliability.  There are two sub-layers of the data link layer they are:

Media Access Control - Only one device at a time can transmit on any media.  If there is more than one they will scramble signal.  There are three ways to control access to media, they are:
Addressing - By physical address??  If you have information please forward it to me.
Contention - With this method anybody can transport any time they want.  Modern contention-based networks listen to the network before they transmit to keep collisions down.  The two ways to do this are:
CSMA/CD - This method is used on Ethernet networks.  It listens and if media available transmits data.  If there is a collision the machine will send out a message to everyone saying to wait a moment to clear the line.  Then everyone will retransmit at a random amount of time.
CSMS/CA - This method is used on AppleTalk networks.  The machine listens to media and if available send a request to send  to the server.  The server responds with a clear to send and the data is transmitted.
Advantages of contention-based system:
Low overhead
High speed on networks with less than roughly 40 percent utilization.
Disadvantages of contention-based networks:
Degradation of performance due to collisions under moderate-to-high network loads
Inability to assign priorities to special devices
Channel access not always predictable
Deterministic - This where a transmitting order is set.  Two types are token passing and polling.
Token passing - a small data frame is passed from device to devices across the network in a predetermined order.
Advantages - 
Special devices can have higher priorities than normal devices.
Token passing is much more efficient under high network loads than contention-based networks.
Network access is predictable due to the predetermined transmitting order.
Disadvantages - 
Contention-based networks with low utilization are faster.
Hardware more expensive.
Polling systems - a master device checks secondary devices to see if they need to transmit.
Advantages - 
The priority and amount of data allowed to be transmitted at a time can be predetermined.
Little bandwidth is lost when the network reaches high utilization.
Disadvantages -
The process of polling each secondary device uses more bandwidth than the other methods.
Transmissions can be delayed from secondary devices while they wait to be polled by the primary device.
Logical Link Control - Maintains data link connections between network devices.  It is responsible for any flow control and error correction found in this layer.  It implements flow control by one of two methods:
Sliding windows:  Using this the receiving device can send one acknowledgement for a group of frames.
Stop and wait:  Sends frames until memory is full then waits for an acknowledgement for more.

 

 

Physical Layer - The function of this layer is the transmission of bits over the network media.  It also specifies requirements for how transmission occurs and ensures compatible data transmission with other devices.