Layer 5: The Session Layer

The fifth layer of the OSI Model is the Session Layer. This layer is relatively unused many protocols bundle this layer’s functionality into their transport layers.

The function of the OSI Session Layer is to manage the flow of communications during a connection between two computer systems. This flow of communications is known as a session. It determines whether communications can be uni- or bi-directional. It also ensures that one request is completed before a new one is accepted.
The Session Layer also may provide some of the following enhancements:
• Dialog control
• Token management
• Activity management

A session, in general, allows two-way communications (full duplex) across a connection. Some applications may require alternate one-way communications (half duplex). The Session Layer has the option of providing two-way or one-way communications: an option called dialog control.

For some protocols, it is essential that only one side attempt a critical operation at a time. To prevent both sides from attempting the same operation, a control mechanism, such as the use of tokens, must be implemented. When using the token method, only the side holding a token is permitted to perform the operation. Determining which side has the token and how it is transferred between the two sides is known as token management. The token management functions are used in the ISO session layer protocols.

The use of the word “token” here should not be confused with Token Ring operation. Token management is a much higher level concept at Layer 5 of the OSI model. IBM’s Token Ring operation belongs to Layers 2 and 1 of the OSI model.

If you are performing a one-hour file transfer between two machines, and a network crash occurs approximately every 30 minutes, you might never be able to complete the file transfer. After each transfer aborts, you have to start all over again. To avoid this problem, you can treat the entire file transfer as a single activity with checkpoints inserted into the datastream. That way, if a crash occurs, the session layer can synchronize to a previous checkpoint. These checkpoints are called synchronization points.

There are two types of synchronization points: major and minor synchronization points. A major synchronized point inserted by any communicating side must be acknowledged by the other communicating side, whereas a minor synchronization point is not acknowledged. That portion of the session that is between two major synchronization points is called a dialog unit. The operation of managing an entire activity is called activity management. An activity can consist of one or more dialog units.

TCP/IP networks do not have a general Session Layer protocol. This is because some of the characteristics of the Session layer are provided by the TCP protocol. If TCP/IP applications require special session services they provide their own. An example of such a TCP/IP application service is the Network File System (NFS), which implements its own Session Layer service the Remote Procedure Call (RPC) protocol.

The Network File System (NFS) protocol also uses its own Session Layer service: the External Data Representation (XDR) protocol.