1.2LANs, Ethernet, and switching

In office and data-center LAN design, the choice of "hub vs. switch" and "is an L2 switch enough, or is an L3 switch needed" greatly affects performance, cost, and scalability. This section builds on understanding why CSMA/CD is no longer a concern in modern switched LANs and why MAC address learning allows communication to work even across segmented ports, to develop the judgment criteria used for real-world device selection.

1.2.1Ethernet standards, CSMA/CD, and full duplex

• Ethernet is the LAN standard defined by IEEE 802.3. Standards are named by speed, such as 1000BASE-T (1 Gbps) or 10GBASE-T (10 Gbps), and further subdivided by medium (twisted pair/optical fiber) and reach.
• CSMA/CD (Carrier Sense Multiple Access with Collision Detection) is a control method for shared media (bus topologies or hub-based half-duplex LANs): when multiple stations transmit simultaneously, a collision occurs, and upon detecting it, stations stop transmitting and retransmit after a random backoff. With full-duplex communication (sending and receiving simultaneously on separate paths) over a switched connection, collisions cannot occur in the first place, so CSMA/CD becomes effectively unnecessary.

1.2.2MAC addresses and L2 switch learning/flooding

• A MAC address is a 48-bit physical address assigned to a network interface (e.g., 00-1A-2B-3C-4D-5E). The upper 24 bits are the vendor identifier (OUI), and the lower 24 bits are a manufacturer-assigned serial number, making the address globally unique in principle.
• An L2 switch records the mapping between a received frame's source MAC address and the receiving port in its MAC address table (learning), and thereafter forwards frames destined for that MAC address only out the matching port. If the destination MAC has not yet been learned, or the destination is a broadcast, the switch performs flooding, sending the frame out every port except the one it arrived on.

Suppose you are the network builder designing the LAN for a newly opened office, with a requirement to accommodate 30 PCs on one floor. The first candidate, a hub (repeater hub), is a physical-layer relay device where all ports share a single collision domain; as the number of connected devices grows, collisions become frequent and effective throughput drops, so it is inadequate on performance grounds at a scale of 30 devices. Considering an L2 switch instead, each port becomes an independent collision domain (full-duplex communication is possible per port), and MAC address learning forwards frames only to the necessary destination, so collisions are effectively eliminated compared to a hub and performance improves substantially. However, all ports still share a single broadcast domain, so broadcasts such as ARP requests still reach every PC. If an additional requirement then emerges—"limit the scope of broadcasts by segmenting the floor per department" and "control inter-department traffic via routing"—you would need an L3 switch (a switch that can carve out VLANs and route between them). If the requirement is simply "connect 30 devices on one floor," an L2 switch satisfies it, but whether segmentation and routing are required is the deciding factor in device selection.

1.2.3Section summary

• Full-duplex communication eliminates collisions, making CSMA/CD unnecessary. A hub is half-duplex with a single collision domain shared across all ports
• An L2 switch learns the source MAC plus receiving port, and floods frames when the destination is unlearned or a broadcast
• An L2 switch segments only the collision domain; segmenting the broadcast domain requires a VLAN-capable L3 switch