TCP/IP - model, encapsulation, packet vs. frame

Internet protocol suite, commonly known as TCP/IP, is a set of communication protocols, where the main ones are the Transmission Control Protocol (TCP) and Internet Protocol (IP).

TCP/IP model

The classic four-layer TCP/IP model, sometimes also called the Internet Reference Model, is based on the OSI model, but is closer to practice.

Sometimes the lowest layer is divided into two parts. The physical layer (Physical Layer) and the data link layer (Data link).

Encapsulation when sending data

When sending data, encapsulation (wrapping) is performed from the highest layer down. It happens like this: the application takes the data (application layer) it wants to send to another station and adds an application header. This data is sent to the lower layer (transport), which divides the data being sent into segments, wraps them, and adds a TCP (or UDP) header to create a TCP segment. The next layer (network) adds an IP header, and this creates an IP packet (sometimes referred to as an IP datagram). In the last layer (access), an Ethernet header is added to the beginning of the packet and a trailer to the end, which contains the FCS (Frame Check Sequence) - checksum (often CRC - cyclic redundancy check is used for its calculation). This is how an Ethernet frame is created in the last step, which is transmitted to the communication medium.

Note: The above description concerned the IP protocol, if it is an auxiliary protocol (like ARP, ICMP), the situation is similar.

When the destination device receives the data, the reverse process of de-encapsulation (unwrapping) is performed from the lowest layer up, and the destination application receives the sent data.

Packet vs. frame

In connection with data and its transmission over the network, two important terms are used, these are packet and frame. These terms are widely used in literature, on the internet, and in practice. Unfortunately, there are several different explanations in the literature, and the use of these terms also varies. Moreover, it depends on the technology in which they are used (similar terms are used, for example, in serial communication). I present here the explanation of these terms that seems most correct to me.

Packet means a package in translation and is a formatted block of data that is transmitted in a computer network. Packets are talked about in connection with the network layer. A packet contains an IP address, other attributes, and data. It is wrapped in a frame and then travels through the network.

Frame is what actually travels in the network. Frames are created at the physical layer of the network interface. The name suggests that it is more of a time segment than an object. There are several types of frames, the most used is Ethernet II, which can transport both TCP/IP and IPX/SPX.

The maximum size of data transmitted in a frame (i.e., the size of the packet) is determined by MTU (Maximum transmission unit) and is related to the link layer. The standard MTU size in Ethernet is 46 - 1500 bytes. At the beginning of communication, devices agree on the MTU size using ICMP. If the other side requires a smaller MTU, fragmentation - splitting the frame into several smaller ones - can be performed. Fragmentation can be prohibited by a flag in the packet header. If fragmentation is prohibited and the other side accepts only a smaller MTU, it will respond with an ICMP error.

In general terminology, the term Protocol Data Unit (PDU) is used, which is a data unit at any layer (IP packet, frame, etc.). When a PDU is passed to a lower layer, it is referred to as a Service Data Unit (SDU), where it is supplemented with Protocol Control Information (PCI), i.e., a header, and a new PDU is created.

Packet - frame format

The basic shape of a frame is header, data (payload), and trailer. The data consists of encapsulated data from higher layers. The following figure shows the four layers of the TCP/IP model along with the rough format of the data as it is encapsulated from top to bottom. The example given is for the TCP protocol, but the general model is the same.

Ethernet header

The Ethernet header contains the source and destination MAC address (physical address from hop to hop) and type/length. The type in the header determines which protocol is used at the higher layer (IP, ARP, etc.). To distinguish individual frames, a special introduction (preamble) composed of a sequence of alternating ones and zeros and SFD - start frame delimiter is transmitted at the beginning.

IP header

The IP header contains the source and destination IP address (logical address of the communicating parties), as well as the protocol specification for the next layer (TCP, UDP, ICMP, etc.), header checksum, and other values. The standard size of the IP header is 20B, but it can also contain optional properties. The shape of the IP header depends on the protocol used; the following example is for IPv4.

TCP header

The TCP header contains the source and destination port, sequence and acknowledgment numbers (to ensure reliability), flags (which include SYN and ACK for connection establishment), window size, and more. The standard size is again 20B.