You can think of the data link layer protocol as the fundamental "language" spoken by networks. This protocol must be compatible with the physical cables that are used, but in many cases the physical cables can support a variety of different protocols. Each device on the network speaks a particular data link layer protocol. In the past, there were literally dozens of protocols that were used; each protocol was custom-tailored to specific needs of the devices and application software in use. Where different devices or cables from different parts of the organization were connected, we used a translator to convert from the data link protocol spoken by one device into the protocol spoken by another device.
As the Internet has become more prominent and as it has become more important to move data from one part of an organization to the other, the need to translate among different data link layer protocols has become more and more costly. It is now more important to provide a few widely used protocols for all networks than to custom tailor protocols to the needs of specific devices or applications. Today, businesses are moving rapidly to reduce the number of different protocols spoken by their networking equipment and converge on a few standard protocols that are used widely throughout the network.
We still do use different protocols in different parts of the network where there are important reasons for doing so. For example, local area networks often have different needs than wide area networks, so their data link layer protocols typically are still different, but even here we are seeing a few organizations move to standardize protocols.
This move to standardize data link layer protocols means that networking equipment and networking staff need to understand fewer protocols—their job is becoming simpler, which in turn means that the cost to buy and maintain network equipment and to train networking staff is gradually decreasing (and the side benefit to students is that there are fewer protocols to learn!). The downside, of course, is that some applications may take longer to run over protocols are not perfectly suited to them. As network capacities in the physical layer continue to increase, this has proven to be far less important than the significant cost savings that can be realized from standardization.
SUMMARY
Media Access Control Media access control refers to controlling when computers transmit. There are three basic approaches. With roll-call polling, the server polls client computers to see if they have data to send; computers can transmit only when they have been polled. With hub polling or token passing, the computers themselves manage when they can transmit by passing a token to one other; no computer can transmit unless it has the token. With contention, computers listen and transmit only when no others are transmitting. In general, contention approaches work better for small networks that have low levels of usage, whereas polling approaches work better for networks with high usage.
Sources and Prevention of Error Errors occur in all networks. Errors tend to occur in groups (or bursts) rather than 1 bit at a time. The primary sources of errors are impulse noises (e.g., lightning), cross-talk, echo, and attenuation. Errors can be prevented (or at least reduced) by shielding the cables; moving cables away from sources of noise and power sources; using repeaters (and, to a lesser extent, amplifiers); and improving the quality of the equipment, media, and their connections. Error Detection and Correction All error-detection schemes attach additional error-detection data, based on a mathematical calculation, to the user’s message. The receiver performs the same calculation on incoming messages, and if the results of this calculation do not match the error-detection data on the incoming message, an error has occurred. Parity, checksum, and CRC are the most common error-detection schemes. The most common error-correction technique is simply to ask the sender to retransmit the message until it is received without error. A different approach, forward error correction, includes sufficient information to allow the receiver to correct the error in most cases without asking for a retransmission.
Message Delineation Message delineation means to indicate the start and end of a message. Asynchronous transmission uses start and stop bits on each letter to mark where they begin and end. Synchronous techniques (e.g., SDLC, HDLC, Ethernet, PPP) group blocks of data together into frames that use special characters or bit patterns to mark the start and end of entire messages. Transmission Efficiency and Throughput Every protocol adds additional bits to the user’s message before sending it (e.g., for error detection). These bits are called overhead bits because they add no value to the user; they simply ensure correct data transfer. The efficiency of a transmission protocol is the number of information bits sent by the user divided by the total number of bits transferred (information bits plus overhead bits). Synchronous transmission provides greater efficiency than does asynchronous transmission. In general, protocols with larger frame sizes provide greater efficiency than do those with small frame sizes. The drawback to large frame sizes is that they are more likely to be affected by errors and thus require more retransmission. Small frame sizes are therefore better suited to error-prone circuits, and large frames, to error-free circuits.
