Telecommunication protocols: OSI model (continued)

In the previous part (link), we divided all the layers of the seven-layer OSI model into two groups. We examined the first four layers that form the 'Host Layers' group. In this post, we will look closer at the second, or 'Media Layers' group, which includes the three lower levels of the network model.

The 'Network Layer' routes blocks of data (packets or messages) through complex composite networks. To solve this primary task, the network layer protocol must define the structure of network layer addresses, the format of data blocks, and the type of services (with or without establishing the connection) it provides to the underlying transport layer. The protocol must have built-in mechanisms for requesting and ensuring the required quality of services offered and functions for managing the congestion of the communication channel.

The network layer's complexity determines the transport layer's ease of implementation. If the network layer protocol is connection-oriented, then there is no need to build into the transport layer mechanisms for ordering data blocks and error control, which means it is simpler.

However, connection-oriented network protocols must include descriptions of the procedures for establishing, maintaining, and terminating a connection instead of connectionless protocols. This complexity pays off because connection establishment provides greater control over its quality.

The type of network protocol used depends on the type of network traffic. Network traffic can be divided into two types: error-sensitive and latency-tolerant and latency-sensitive and error-tolerant. The first type includes almost all general-purpose data, and the second type includes all voice and video communications.

In connection-oriented networks, nodes forward messages according to the virtual connection number. Moreover, messages (with the exception of those sent when setting up a logical connection) do not contain information about the connection's final destination.

Each packet contains all the necessary information about its destination in connectionless networks. The router decides where to forward this packet next. The routing process consists of two operations: building a routing table and forwarding individual packets.

Today's most popular one is the 'Internet Protocol' (IP). In conditions of traffic convergence, when IP protocol packets must carry multimedia data in one stream (voice and video over IP technologies), its most significant drawback is the lack of traffic prioritization. Therefore, the primary efforts in developing the IP protocol aim to introduce an appropriate mechanism to give it some typical properties of a connection-oriented protocol, allowing it to successfully transport both main types of network traffic. It is proposed that the 'Multiprotocol Label Switching' (MPLS) method be used to adapt the IP protocol to modern requirements.

The 'Data Link' layer must ensure error-free data delivery to the network layer. Thus, its primary purpose is to identify and eliminate physical layer errors. Unlike the one at the physical level, information at the link level is transmitted in a structured form, i.e., frames. How the link layer corrects errors depends on the type of physical link. Direct correction is based on additional information in the frame. This method is usually used for transmission channels with numerous errors (subscriber network access lines) or high latency (satellite channels). Some modern link-layer protocols do not correct the errors per se but just detect and discard corrupted frames. Errors are resolved with higher-level protocols. Protocols that use this method are designed for high-quality physical channels with a low probability of errors, particularly fiber optic cables.

The data link layer must provide the following services:

• communication channel access control, determining the order of units' frame transmission;
• frame synchronization, which sets the beginning and end of each frame;
• data delimitation, when the receiving device separates the data and control information that the link layer adds to the transmitted frame;
• error detection and data recovery. The most common method for finding errors in the 'Cyclic Redundancy Check' (CRC);
• data flow control, since the receiving device must be able to stop the transmission if it is busy. Naturally, after freeing up the resources needed, the receiving device must be able to signal again that it is ready to continue receiving data;
• addressing for multipoint connections with more than two devices to indicate the recipient and identify the sender;
• opening and closing a connection.

It should be noted that as the quality of the physical layer improves, the need for all link layer functions decreases. In these circumstances, it is reasonable to leave the control of late-frame delivery to the end nodes. Restoring lost frames can be assigned to higher layers (for example, the transport one).

Perhaps the most common direct link layer protocol is the 'Point-to-Point Protocol' (PPP). This protocol for interacting with peer-to-peer systems involves a full-duplex physical channel, dedicated or switched. It is interesting that considering the new realities of the physical layer, it only provides error checking by the CRC procedure but does not include error correction. The PPP protocol simply drops erroneous frames and does not deal with frame ordering, relying on the physical layer.

The 'Physical' layer carries out the transfer of data bits. Its main difference from the other OSI model layers is that it is both necessary and sufficient for the exchange of data between two points.

The physical level is characterized by several important features.

• The physical layer architecture provides simplex, half-duplex, and full-duplex schemes for device interactions. In addition to the point-to-point topology, a point-to-multipoint topology is possible, particularly "star" and "bus." The "ring" topology is also very common; it is a closed structure of many point-to-point connections. In addition, the physical layer architecture should include parallel and serial transmission schemes. In a parallel scheme, many identical communication channels are organized between interacting devices, one of which is reserved for synchronization. Thus, the total throughput of the connection is equal to the sum of the throughputs of all communication channels used for transmitting data. In practice, the parallel transmission range is limited by the delay time skew of the component communication channels.
• The data transmission environment can include twisted pair cable, coaxial cable, fiber optic cable, radio waves, line-of-sight wireless optical transmission, and an AC network. These all differ in throughput, noise immunity, installation features, and spread.
• Signal transmission includes transporting analog signals through digital systems (for example, PCM systems) or digital signals through analog systems (telephone line modems).
• Synchronization is necessary for serial transmission. The data signal is usually represented as a non-return to zero level (NRZ-L) linear code, where a binary one corresponds to a positive polarity pulse with a flat top and an amplitude of V volts, and a binary zero corresponds to a pulse of the same shape with zero amplitude. The NRZ-L code has two downsides: the presence of a DC component and its inability to transmit a sync signal along with the data. Thus, the NRZ-L signal, a sequence of ones, cannot be transmitted by a natural communication line containing capacitors and transformers. Synchronization of the reception signal is possible only with separate transmission of the digital signal's states (zeros and ones). Obviously, with a long sequence of ones, such a separation will not be possible, and, therefore, a separate transmission of the clock signal is needed. The problem is solved by converting the NRZ-L binary code for transmission over the line, for example, into a pseudo-ternary code like AMI or HDB-3.
• There are two types of multiplexing: frequency-division multiplexing (FDM) and time-division multiplexing (TDM). Both allow saving the bandwidth of physical lines, unlike inverse multiplexing, which, on the contrary, is based on using redundant physical lines where possible.

The DTE-DCE and DSU-CSU interfaces are two popular physical layer interfaces.