Switching in Telecommunications
Switching in Data Communications
This chapter discusses switching techniques that have been employed in data networks. Following topics are discussed
In circuit switched networks, each connection between the calling and the called subscribers results in a physical communication channel being setup through the network for the duration of the call. Once this channel has been established, data may flow through it. The connection is terminated at the end of the session and all the resources of the connection are released for use by other connections.
The call setup procedure in circuit switching networks may require a significant amount of time. However, once the connection has been established, network resources for the connection are reserved only for that connection. Thus there is no danger of congestion. Moreover, the two nodes must transmit and receive data at the same rate.
Circuit switching is, therefore, a good choice in environments where significant amount of data needs to be transferred at a fixed rate. However, circuit switching utilises network resources inefficiently as there may be no communication when the two nodes are busy processing previous messages. As a result, the network resources reserved for the connection are not utilised for that period.
In message switching networks, no physical channel is established in advance between the transmitter and the receiver. Instead, the block of data transmitted by the transmitter is stored in the first switching node and then forwarded later, one hop at a time. Each block is received in its entirety, inspected for errors and then transmitted to the next switching node. Networks using such an approach are known as the store and forward networks.
With message switching, there is no limit on the block size. This means that the switches must have disks to buffer large blocks of data. Moreover, a large block may tie up a link between two switches for several minutes. These factors render message switching useless for interactive traffic.
As opposed to message switching, packet switching networks place a tight upper limit on the block size, allowing the packets to be buffered in the switch memory (RAM) instead of the disks. Messages of arbitrarily long lengths are fragmented by the transport layer at the source computer prior to transmission. The transport layer at the destination computer reassembles the appropriate packets to form a single message. The lengths and structure of the packets are defined to facilitate efficient communication.
In packet switched networks, the source computer puts the network addresses of the source and destination computers in the packet header when it creates the packet. The source computer then passes these packets to the local packet switch. This switch stores the packet and then inspects the destination address contained in its header. Each packet switch contains a routing directory specifying the outgoing link to be used for each network address. After determining an appropriate link to forward the packet on towards its destination, the switch forwards the packet to the next packet switch. Intermediate switches on the route perform the same operation on the packet (i.e., store packet, look up destination address, forward packet). Finally the switch at the destination forwards the packet to the destination computer.
Figure : Packet Switched Network
Packet switching networks can either be connection oriented (virtual circuit) or connectionless (datagram). If a connection-oriented paradigm of communication is chosen, the source computer requests the network to form a connection to the destination computer. The destination computer agrees to accept the connection. When both the computers have agreed to communicate, the network establishes a connection and informs both the computers. These computers can then exchange data.
Connection oriented services provide a stream interface for the computers to communicate. The source computer emits a stream of data that is divided into packets by the network for transport. Communication through the connection need not be continuous. Computers may pause as they process data for communication and resume communication when the data is ready. They, however, remain connected unless the computers decide to terminate the connection.
Whenever a computer wishes to transmit data using a connectionless service, it simply places the data in appropriate frame format, attaches the network address of the destination computer and then passes the frame to the network for delivery. The connectionless network system transports the frame to the prescribed destination and delivers it.
Thus, connection oriented services offer a reliable delivery mechanism. However, connection oriented services usually have significantly large overheads attributed to connection establishment, acknowledgements and error control.
Virtual circuits can be of two types, Permanent Virtual Circuits (PVCs) and Switched Virtual Circuits (SVCs). A PVC is like a leased line. It is established by configuring the network switches for a significantly long duration. These are normally used to connect computers that need to transfer large amounts of data frequently. A major advantage of PVCs is their persistence and availability. The configuration of PVCs is generally stored in disks inside the switches. This allows the switches to re-establish connections automatically after a power failure.
Alternatively, a switched virtual circuit is established when the computers wish to communicate. This connection is temporary as the computers request the network to terminate it at the end of the session. Although it is possible to keep a switched connection in place for a long time, most switched connections persist for a short duration. Any computer that uses SVCs to communicate, must re-establish the connection after rebooting from a crash. A major advantage of SVCs is their flexibility and generality. These are useful in situations where infrequent data transfer occurs between different computers.
Cell Switching in ATM Networks
Asynchronous Transfer Mode (ATM) is a packet oriented transfer scheme that is based on short, fixed length packets called cells. ATM cells use a 5 byte header and a 48 byte payload. Short, fixed length cells cause shorter delays and less jitter (or delay variation) than longer and variable length packets. This makes them more suitable for applications using voice and video services. The cell size is fixed to simplify switch design by minimising the overhead required to process each cell.
ATM is connection oriented, i.e. a virtual circuit needs to be established between the source and destination computers before data can flow between them. When the connection is established, the identification details of the virtual circuit are programmed into the switches. These details form a part of the cell header. A cell only contains these identifiers and not the complete source and destination addresses of the computers. Thus, a switch can quickly determine the route that the cell has to take and therefore, switches the cell to its destination. Moreover, once the connection is established, the path is fixed. Every cell using the connection has to take this path in situations other than equipment failure. Thus cells arrive in sequence, a quality considered desirable for isochronous (time dependant) traffic. The quality of service of the connection is also agreed during the connection establishment phase. This relates to traffic requirements (bandwidth, jitter, transmission delay etc.). Thus, ATM networks can endure different types of traffic as they can guarantee the expected quality of connection to the application. Finally, flow control and error handling mechanisms are not carried out in ATM networks but are left to the end user applications or the access devices.
Thus, prior switch configuration enables the implementation of simple routing tables, and small fixed length packets optimise switch design. These features allow switching to be implemented completely in hardware.
The association of a specific quality of service with each virtual circuit is one of the most attractive features of ATM networks. Currently, five different qualities of service have been defined. These qualities of service range from fulfilling the requirements of constant bit rate applications such as voice to high speed, bursty applications such as file transfers between a server and a client. The users can get a data rate across the network that is appropriate to their requirements by specifying a connection with a suitable quality of service. Thus applications that need to relay time critical information (e.g. voice, video etc.) can demand a circuit with transmission parameters that will guarantee that voice and video are delivered in a timely manner with minimal data jitter and low transmission delay. The non-critical data (e.g. e-mail, file transfer etc.), on the other hand, can be delayed inside the network while the physical links are used by the critical traffic.
Previous : A Generic Switch Architecture
Next : Functions Performed by ATM Switches
Top : Contents
© Omar Bashir, December 1998