For many applications the cost of the communication lines exceeds the cost of the equipment connected by those lines. In an attempt to reduce communication costs, many networks provide a way for multiple terminals to share a single communication line. Instead of each terminal being connected directly to the central computer by a separate line, a number of terminals share a line. in this way the number of lines (and hence modems or line drivers) is greately reduced. The conceptual model is that of Fig 1, in which a terminal controller accepts input from a cluster of terminals, and funnels the output onto one line, as well as the reverse operation. In Fig 1(a), all the terminals are wired onto the same multidrop line, whereas in Fig 1(b)(b) each terminal has its own point to point lineto the controller.
With only one line for a group of terminals only one message can be sent at a time, either by a terminal or by a central computer. all transmissions on each line are therefore controlled by the cental computer using the technique known as polling. Such architectures are normally used in applications that involve a single master computer communicating with a distributed community of slave computers. Examples are a back-of-store computer controlling a distributed set of point-of-sale terminals in a department store. All transmissions are between the master and a selected slave computer, so the master controlls the order of all transmisions.
To ensure that only one message is transmitted at any instant on each shared communication line, the central computer ,or its agent (terminal controller), either polls or selects each terminal connceted to the line in a particular sequence. As each terminal connected to the shared line is allocated a unique terminal identifier that the central computer communicates with. Messages can be of two types: control or data. This type of polling known as roll call polling and consists of the controller simply sending a message to each terminal in turn, inquiring whether or not the terminal has anything to say. On half duplex lines each poll requires two line turnarounds, one to allow the controller to send, and one to alow the terminal to send. It may take a long time to complete a cycle on a line with many terminals, even if most are idle most of the time. To overcome this problems, a more common type of multidrop network uses a cluster controller to reduce the responce time of the network. If terminal clusters are widely separated, an alternative mechanism known as hub polling is sometimes cost justified. With hub polling the controller polls the farthest terminal from it. The terminal sends, if it has,data back to the controller else it sends a polling message to its neighbor (on the controller side). The poll propogates from terminal to terminal until one terminal has data to send or until it gets back to the controller.
The central computer ,or its agent (terminal control), either polls or selects each terminal connceted to the line. The controller polls each terminal in turn, inquiring whether or not the terminal has anything to say. If the polled terminal has data to send, it sends the data. If not, it sends back a special "poll reject" message. the central computer than continues by polling or selecting the next terminal. This type of polling results in quite long response times for larger networks since each terminal in the network must be polled or selected before it can send or receive a message. The communication overheads imposed on the central computer can be very high. Another disadvantage is the potential number of negetive responses to polls, which can consume precious resources on the channel.
The poll control message is used to request a spesific terminal to send any waiting data message it may have.Periodically the central computer sends each terminal , in turn , a poll control message , which effectively invites the polled terminal to send a message should it have one waiting : If it has, it is returned in a data message , otherwise it responds with a nothing-to-send message . Usually, the controller just polls all the terminals in round-robin fashion, but in some circumstances important terminals may get several pollings per cycle. The poll protocol operation starts with the central computer first sending an ENQ poll control message with the address of the polled terminal proceding the ENQ character. Then, assuming the polled station has a message awaiting transmission, it responds by sending the message. on receipt of the data block the central computer recomputes the parity check sequence and assuming no transmissions errors acknowledges its correct receipt by sending an ACK if the data is acceptable or a NAK if they are incorrect. This two events (data and ACK/NAK) may occure many times until the terminal has no more data to send. Then the terminal must send an indicator that it has completed its transmission and the logical connection is cleared with an EOT control message (End Of Transmission). This procedure is described in Fig. 3. For example consider the X.25 poll example.
The select control message is used to ask the selected terminal whether it is ready to receive a data message. Whenever the central computer wishes to send a message to a terminal, it sends a select control message addressed to a particular terminal. Assuming the selected terminal is able to receive a message. The central computer then sends the data message. Finally, the terminal acknowledges correct receipt of the data message and the central computer continues by either polling or selecting another terminal. The select command is illustrated in Fig. 3. Selects means the central computer has data for the terminal. The ACK to the select means the terminal is ready to receive data. The data is transmited, checked for errors and acknowledged. The process can repeat itself. Eventually , an EOT control indicator is transmitted. In some situations, when selecting a terminal it is not always necessary to wait for an ACK to the ENQ control message. For example, if a terminal has been selected previously and the logic connection has not been closed, then the central computer sends the message immediately after the select control message, without waiting for an ACK(or NAK) response.
In this configuration each cluster controller is connected to its nearest neighbor rather than to the central computer. The central computer manages all transfers to and from the cluster controllers. The central computer selects and sends a data mesage to any of the controllers at any time using special select lines. To receive messages from the controllers, the central computer sends a poll control message to the furthest controller which responds by sending either a data message or a nothing-to-send control message on another special poll (return) lines to its nearest-neighbour controller. On receipt of this message, the next controller interprets this as a poll message and, if it has a message waiting, responds by adding its own data message to the tail of the received message from its upstream neighbour. The composite message is then forwarded to its downstream neighbour , again on the return line. This procedure continues down the chain, each controller adding its own response message as it relays the message towards the central computer. Finally, on receipt of the composite response message, the terminal controller disassembles the message and passes on any valid data messages contained within it to the central computer for further processing. The advantage here is that it is not necessary on half duplex lines to keep turning the line just to discover that the terminal has nothing to say.
For the case of the star(point-to-point) controller, polling is not required to avoid chaos on the lines. Nevertheless, roll call polling is often used anyway, to allow the master to aquire input in an orderly fashion. This poll messages differ from those of the multidrop lines case because there are no site adresses needed. A terminal only receives those polls directed to it.
For many applications the cost of the communication lines exceeds the cost of the equipment connected by those lines. In an attempt to reduce communication costs, many networks provide a way for multiple terminals to share a single communication line. Instead of each terminal being connected directly to the central computer by a separate line, a number of terminals share a line. in this way the number of lines (and hence modems or line drivers) is greately reduced. The conceptual model is that of Fig 1, in which a terminal controller accepts input from a cluster of terminals, and funnels the output onto one line, as well as the reverse operation. In Fig 1(a), all the terminals are wired onto the same multidrop line, whereas in Fig 1(b)(b) each terminal has its own point to point lineto the controller.
With only one line for a group of terminals only one message can be sent at a time, either by a terminal or by a central computer. all transmissions on each line are therefore controlled by the cental computer using the technique known as polling. Such architectures are normally used in applications that involve a single master computer communicating with a distributed community of slave computers. Examples are a back-of-store computer controlling a distributed set of point-of-sale terminals in a department store. All transmissions are between the master and a selected slave computer, so the master controlls the order of all transmisions.
To ensure that only one message is transmitted at any instant on each shared communication line, the central computer ,or its agent (terminal controller), either polls or selects each terminal connceted to the line in a particular sequence. As each terminal connected to the shared line is allocated a unique terminal identifier that the central computer communicates with. Messages can be of two types: control or data. This type of polling known as roll call polling and consists of the controller simply sending a message to each terminal in turn, inquiring whether or not the terminal has anything to say. On half duplex lines each poll requires two line turnarounds, one to allow the controller to send, and one to alow the terminal to send. It may take a long time to complete a cycle on a line with many terminals, even if most are idle most of the time. To overcome this problems, a more common type of multidrop network uses a cluster controller to reduce the responce time of the network. If terminal clusters are widely separated, an alternative mechanism known as hub polling is sometimes cost justified. With hub polling the controller polls the farthest terminal from it. The terminal sends, if it has,data back to the controller else it sends a polling message to its neighbor (on the controller side). The poll propogates from terminal to terminal until one terminal has data to send or until it gets back to the controller.
The central computer ,or its agent (terminal control), either polls or selects each terminal connceted to the line. The controller polls each terminal in turn, inquiring whether or not the terminal has anything to say. If the polled terminal has data to send, it sends the data. If not, it sends back a special "poll reject" message. the central computer than continues by polling or selecting the next terminal. This type of polling results in quite long response times for larger networks since each terminal in the network must be polled or selected before it can send or receive a message. The communication overheads imposed on the central computer can be very high. Another disadvantage is the potential number of negetive responses to polls, which can consume precious resources on the channel.
The poll control message is used to request a spesific terminal to send any waiting data message it may have.Periodically the central computer sends each terminal , in turn , a poll control message , which effectively invites the polled terminal to send a message should it have one waiting : If it has, it is returned in a data message , otherwise it responds with a nothing-to-send message . Usually, the controller just polls all the terminals in round-robin fashion, but in some circumstances important terminals may get several pollings per cycle. The poll protocol operation starts with the central computer first sending an ENQ poll control message with the address of the polled terminal proceding the ENQ character. Then, assuming the polled station has a message awaiting transmission, it responds by sending the message. on receipt of the data block the central computer recomputes the parity check sequence and assuming no transmissions errors acknowledges its correct receipt by sending an ACK if the data is acceptable or a NAK if they are incorrect. This two events (data and ACK/NAK) may occure many times until the terminal has no more data to send. Then the terminal must send an indicator that it has completed its transmission and the logical connection is cleared with an EOT control message (End Of Transmission). This procedure is described in Fig. 3. For example consider the X.25 poll example.
The select control message is used to ask the selected terminal whether it is ready to receive a data message. Whenever the central computer wishes to send a message to a terminal, it sends a select control message addressed to a particular terminal. Assuming the selected terminal is able to receive a message. The central computer then sends the data message. Finally, the terminal acknowledges correct receipt of the data message and the central computer continues by either polling or selecting another terminal. The select command is illustrated in Fig. 3. Selects means the central computer has data for the terminal. The ACK to the select means the terminal is ready to receive data. The data is transmited, checked for errors and acknowledged. The process can repeat itself. Eventually , an EOT control indicator is transmitted. In some situations, when selecting a terminal it is not always necessary to wait for an ACK to the ENQ control message. For example, if a terminal has been selected previously and the logic connection has not been closed, then the central computer sends the message immediately after the select control message, without waiting for an ACK(or NAK) response.
In this configuration each cluster controller is connected to its nearest neighbor rather than to the central computer. The central computer manages all transfers to and from the cluster controllers. The central computer selects and sends a data mesage to any of the controllers at any time using special select lines. To receive messages from the controllers, the central computer sends a poll control message to the furthest controller which responds by sending either a data message or a nothing-to-send control message on another special poll (return) lines to its nearest-neighbour controller. On receipt of this message, the next controller interprets this as a poll message and, if it has a message waiting, responds by adding its own data message to the tail of the received message from its upstream neighbour. The composite message is then forwarded to its downstream neighbour , again on the return line. This procedure continues down the chain, each controller adding its own response message as it relays the message towards the central computer. Finally, on receipt of the composite response message, the terminal controller disassembles the message and passes on any valid data messages contained within it to the central computer for further processing. The advantage here is that it is not necessary on half duplex lines to keep turning the line just to discover that the terminal has nothing to say.
*This article information is the property of Rad Datra Communications
Pulse Supply has products and solutions for all these technologies.
Please call us for a free consultation.
