Technical definition
The OSI Data Link layer is layer 2 of the OSI reference model. It transfers link-layer protocol data units between adjacent network entities over a physical connection, using link-protocol mechanisms such as framing, access control and error detection.
In plain English: Carries data as frames across a link. Ethernet uses MAC addresses for local delivery and an error check to detect damaged frames.
What this layer does
An Ethernet MAC address has 48 bits, usually written as six pairs of hexadecimal digits, such as 02:00:00:00:00:20. Hexadecimal uses 0–9 and A–F. MAC addresses can be locally assigned or changed; they are not guaranteed permanent, globally unique device identities.
A switch learns source MAC addresses on incoming ports and looks up destination MAC addresses to forward frames. An unknown unicast destination is normally flooded within its VLAN through eligible ports except the incoming port. The frame check sequence (FCS) detects corruption; Ethernet does not repair the damaged payload.
A switch forwards within the local network using MAC addresses; it does not choose the IP route to a remote network.
Follow one request from PC-1
PC-1 is a user's computer opening a page on the lab web server. Its IPv4 address is 192.168.1.20/24 and its gateway is R1 at 192.168.1.1. The server is on another subnet, at 203.0.113.20/24, with gateway 203.0.113.1. The /24 means that the first 24 bits identify each subnet. R1 joins the two subnets.
On small screens, swipe sideways or focus the diagram and use the arrow keys.
In this request: PC-1 puts the remote-server packet inside an Ethernet frame addressed to R1's Gi0/0 MAC. R1 routes it and builds another frame for its Gi0/1 link to the server.
The interface labels identify the ends of the cables: Eth0 is the example Ethernet interface on each host, while Gi0/0 and Gi0/1 are R1's two Gigabit Ethernet interfaces. Actual device names vary. An IP address identifies an interface; TCP port 443 identifies the HTTPS service endpoint. It is not a physical socket on the router.
Where it fits in the other model
The four-layer TCP/IP model groups Physical and Data Link functions into its Link or Network Access layer. A common five-layer teaching view separates them again.
The trade-off is between a more detailed vocabulary and a simpler description of deployed protocols. OSI names seven responsibilities; the four-layer TCP/IP model groups some of them together. Neither view requires a separate header for every named responsibility.
When it does not work
If the gateway cannot be resolved, check the active interface, its VLAN and the gateway IP before assuming an application failure. An incomplete ARP entry does not establish why resolution failed. On a switch, a wrong VLAN or a blocked link can prevent local delivery. Growing FCS error counters can point to link or hardware problems; Ethernet detects corrupt frames rather than repairing their contents.
Try a practical check
Use PowerShell on Windows for this observation exercise. These commands inspect your own connection or request a public example service; they do not configure R1 or contact the documentation server in the figure.
Get-NetNeighbor -AddressFamily IPv4
arp -aOn Windows PowerShell, compare IP and link-layer address entries on your active interface. Reachable, stale and incomplete entries describe different cache conditions; a cached MAC does not prove that the peer is responding now. For the diagram's remote destination, the first-hop Ethernet frame uses the local gateway MAC. The actual gateway and MAC on your computer will differ.
Your task: identify one useful field in the result, state what it shows, then name something it cannot prove. If the command is unavailable or a request is blocked, record that limitation rather than treating it as proof of a failed layer.
Check your understanding
PC-1 sends to a server on another subnet. Which destination MAC belongs in its first Ethernet frame?