The client –server model distributes the processing between user A’s (client) computer and the central file server. Both computers are part of the network, but each is assigned functions that it performs best. For example, the record – searching portion of an application is placed at the server, and the data – manipulation portion is on the client computer. Thus, only a single record, rather than the entire file, must be locked and sent to the client for processing. After processing, the record is returned to the server, which restores it to the table and removes the lock. This approach reduces traffic and allows more efficient use of shared data. Distributing the record – searching logic of the client’s application to the server permits other clients to access different records in the same file simultaneously. The client – sever approach can be applied to any topology (for example, ring, star, or bus). Figure 3.23 illustrates the client – server model applied to a bus topology.
NETWORK CONTROL
In this section, we examine methods for controlling communications between the physical devices connected to the network. Network control exists at several points in the network architecture. The majority of network control resides with software in the host computer, but control also resides in servers and terminals at the nodes and in switches located throughout the network. The purpose of network control is to perform the following tasks:
1. Establish a communications session between the sender and the receiver.
2. manage the flow of data across the network.
3. Detect and resolve data collisions between competing nodes.
4. Detect errors in data that line failure or signal degeneration cause.
Data Collision
To achieve effective network control, there must be an exclusive link or session established between a transmitting and a receiving node. Only on node at a time can transmit a message on a single line. Two or more signals transmitted simultaneously will result in data collision, which destroys both messages. When this happens, the messages must be retransmitted. There are several techniques for managing sessions and controlling data collisions, but most of them are variants of three basic methods : polling, token passing, and carrier sensing.
Polling
Polling is the most popular technique for establishing a communication session in WANs. One site, designated the master, polls the other slave sites to determine if they have data to transmit. If a slave responds in the affirmative, the master site locks the network while the data are transmitted. The remaining sites must wait until they are polled before they can transmit. The polling technique illustrated in Figure 3.24 is well suited to both the star and the hierarchical topologies. There are two primary advantages to polling. First, polling is non contentious. Because nodes can send data only when the master node requests, two nodes can never access the network at the same time. Data collisions are, therefore, prevented. Second, an organization can set priorities for data communications across the network. Important nodes can be polled more often than less important nodes.
หน้า 119
Token Passing
Token passing involves transmitting a special signal-the token-around the network from node to node in a specific sequence. Each node on the network receives the token, regenerates it, and passes it to the next node. Only the node possessing the token is allowed to transmit data.
Token passing can be used with either ring or bus topologies. On a ring topology, the order in which the nodes are physically connected determines the token-passing sequence. With a bus, the sequence is logical, not physical. The token is passed from node to node in predetermined order to form a logical ring. Token bus and token ring configurations are illustrated in Figure 3.25. Because nodes are permitted to transmit only when they possess the token, the node wishing to send data across the network seizes the token upon receiving it. Holding the token blocks other nodes from Transmitting and ensures that no data collisions will occur. After the transmitting node sends its message and receives an acknowledgment signal from the receiving node, it releases the token. The next node in sequence then has the option of either seizing the t