Demystifying Network Layers: From Physical to Application Explained

Network knowledge is essential for embedded Linux development, yet many struggle with the many layers and their functions. Below is a concise overview of the classic network model.

Physical Layer

The physical layer corresponds to hardware: network interface cards (NICs), PHY chips, and the actual cables connecting PCs, switches, and cloud servers. It handles the transmission of raw bits (0s and 1s) over the medium.

Data Link Layer

This layer introduces the globally unique MAC address, allowing devices to identify each other on the same local network. It also defines how bits are grouped into frames.

Network Layer

Here the IP address comes into play. While MAC addresses are unique, IP addresses may not be; a device can have multiple IPs. The network layer determines routing and subnetting.

IP Protocol

IPv4 uses a 32‑bit address, typically written as four decimal octets (e.g., 172.16.254.1). The address is split into a network part and a host part. Devices in the same subnet share the same network prefix.

To identify whether two IPs belong to the same subnet, a subnet mask is used. The mask is a 32‑bit number with network bits set to 1 and host bits set to 0 (e.g., 255.255.255.0 for a /24 network). By performing a bitwise AND between each IP and the mask, the resulting network addresses can be compared.

Transport Layer

The transport layer provides sockets and protocols such as TCP and UDP. It also introduces ports so multiple applications on the same device can communicate independently.

UDP is simple, adding only source and destination ports, but offers no reliability.

TCP adds reliability through acknowledgments, retransmissions, and ordering, at the cost of increased complexity and resource usage.

Application Layer

Application‑level protocols (e.g., HTTP, FTP, SMTP) encapsulate data for specific services. For example, a web browser uses HTTP over TCP/IP to retrieve web pages.

Data transmission follows a layered encapsulation: each layer adds its header before passing the packet down, and the receiver strips headers layer by layer to retrieve the original payload.