Local Area Networks (LAN)

When you plug a network cable into your computer or connect to a wireless access point, you're utilizing technologies that have been refined over decades of innovation, standardization, and real-world deployment. LAN technologies define the physical and data link layer protocols that enable devices to communicate within a local network.

While many technologies have competed in the LAN space, one has emerged as the overwhelming victor: Ethernet. From its origins at Xerox PARC in 1973 to its current manifestations operating at hundreds of gigabits per second, Ethernet's success story is one of continuous evolution while maintaining backward compatibility—a rare achievement in technology.

However, understanding LAN technologies requires more than just studying Ethernet. The alternatives—Token Ring, FDDI, ATM, and wireless technologies—each contributed concepts and capabilities that influenced modern networking. Some specialized applications still use these technologies today.

Ethernet is a family of wired LAN technologies standardized as IEEE 802.3. It defines physical layer specifications (cables, connectors, signaling) and data link layer protocols (frame format, addressing, error detection). Ethernet's dominance stems from its simplicity, low cost, and remarkable ability to scale from 10 Mbps to 400 Gbps while maintaining fundamental compatibility.

Historical Origins:

Ethernet was invented by Robert Metcalfe and David Boggs at Xerox PARC in 1973. The name derives from 'ether'—the hypothetical medium once thought to permeate space and carry light waves. Just as the ether was imagined as a universal medium for electromagnetic waves, Ethernet was conceived as a universal medium for data communication.

The original Ethernet operated at 2.94 Mbps over coaxial cable using CSMA/CD. It connected Alto computers—early personal workstations that pioneered graphical user interfaces. This proof-of-concept evolved into the commercial standard we use today.

The 10 Mbps Era:

The first commercial Ethernet standard (1980, later IEEE 802.3 in 1983) operated at 10 Mbps. Several physical layer variants emerged:

10BASE-T proved transformative. By using inexpensive twisted-pair cable (the same cable used for telephones) and standardizing on the RJ-45 connector, it made Ethernet accessible to every office. The star topology enabled by 10BASE-T also allowed switches to replace hubs, eventually eliminating the collision domain problem entirely.

The DIX/IEEE 802.3 Distinction:

Two slightly different frame formats exist:

• Ethernet II (DIX): Uses a 2-byte EtherType field to identify the payload protocol (e.g., 0x0800 for IPv4, 0x0806 for ARP)
• IEEE 802.3: Uses a 2-byte Length field plus 802.2 LLC header

In practice, the EtherType field in Ethernet II frames always has values greater than 1500 (the maximum frame length), so devices can distinguish between the two formats by checking if the field value is ≤1500 (length) or >1500 (EtherType).

Understanding the Ethernet frame structure is fundamental to network troubleshooting, protocol analysis, and network programming. Every piece of data transmitted over a LAN is encapsulated in an Ethernet frame.

Complete Ethernet Frame on the Wire:

Detailed Field Analysis:

Preamble and SFD (8 bytes total): The preamble's 7 bytes of 10101010 patterns allow the receiver's clock recovery circuit to synchronize with the incoming signal. The SFD (10101011) marks the end of timing synchronization and the start of the actual frame data. These fields are not considered part of the frame for length calculations.

Destination MAC Address (6 bytes): This identifies the intended recipient(s):

• Unicast: Single recipient (first bit of first byte = 0)
• Multicast: Group of recipients (first bit = 1, but not broadcast)
• Broadcast: All devices (FF:FF:FF:FF:FF:FF)

Switches use this address to determine which port(s) should receive the frame.

Source MAC Address (6 bytes): This identifies the sender. It must always be a unicast address (multicast source addresses are invalid). Switches learn MAC addresses by observing source addresses in incoming frames.

EtherType (2 bytes): Common values include:

• 0x0800: IPv4
• 0x0806: ARP (Address Resolution Protocol)
• 0x86DD: IPv6
• 0x8100: VLAN-tagged frame (802.1Q)
• 0x8847: MPLS unicast

Payload (46-1500 bytes): The minimum payload size of 46 bytes ensures frames are long enough for collision detection to work in classic Ethernet. If the actual data is smaller, padding bytes are added.

The maximum payload of 1500 bytes is the Maximum Transmission Unit (MTU). Larger payloads require fragmentation at higher layers or the use of jumbo frames (up to 9000 bytes in supporting networks).

Frame Check Sequence (4 bytes): A CRC-32 checksum computed over the destination address, source address, EtherType, and payload. The receiver recalculates this value; if it doesn't match, the frame is silently discarded. There is no retransmission mechanism at Layer 2—that responsibility falls to higher layers (e.g., TCP).

Ethernet's success lies in its ability to increase speed by orders of magnitude while maintaining the same frame format and addressing scheme. This backward compatibility protects infrastructure investments and simplifies network design.

Speed Milestones:

Fast Ethernet (100 Mbps):

100BASE-TX became the sweet spot for desktop connectivity through the late 1990s and 2000s. It used the same RJ-45 connectors and Category 5 cabling as 10BASE-T, making upgrades simple. Auto-negotiation allowed 10/100 Mbps devices to coexist on the same network.

Gigabit Ethernet (1 Gbps):

Gigabit Ethernet represented a qualitative shift:

• Full-duplex only for switched environments (half-duplex existed but was rarely used)
• 802.3ab (1000BASE-T) enabled Gigabit over Category 5e/6 copper, using sophisticated signal processing with all four wire pairs
• 802.3z defined fiber variants (1000BASE-SX, 1000BASE-LX) for longer distances

Gigabit Ethernet is now the standard for desktop connectivity in enterprises.

10 Gigabit Ethernet (10 Gbps):

10GbE was designed primarily for backbone and server connectivity:

• Originally fiber-only (10GBASE-SR, 10GBASE-LR)
• 10GBASE-T (802.3an, 2006) brought 10 Gbps to copper, but required Category 6A cabling and consumed more power
• Eliminated CSMA/CD entirely—full-duplex always

25, 40, 100, 400 Gigabit and Beyond:

Modern high-speed Ethernet uses sophisticated techniques:

• Lane aggregation: Multiple parallel channels combined (4x25G = 100G, 4x100G = 400G)
• PAM4 signaling: 4-level pulse amplitude modulation doubles bits per symbol
• Forward Error Correction (FEC): Required at higher speeds to maintain acceptable error rates
• SerDes advancement: Serializer/Deserializer technology improvements

Switches are the fundamental building blocks of modern Ethernet LANs. They replaced hubs in the 1990s and transformed Ethernet from a shared medium into a switched fabric with dedicated bandwidth per port.

How Switches Work:

Switches operate at Layer 2 (Data Link layer), making forwarding decisions based on MAC addresses:

1. Learning: When a frame arrives, the switch records the source MAC address and the port it arrived on in the MAC address table (also called CAM table)

2. Forwarding Decision: The switch looks up the destination MAC address in its table:

   • If found → forward the frame out the specific port
   • If not found → flood the frame to all ports (except the source)
   • If broadcast → flood to all ports in the VLAN

3. Aging: MAC address entries expire after a timeout (typically 300 seconds) if no frames are seen from that address

Dedicated Bandwidth:

In a hub-based network, all devices share the available bandwidth. With 10 devices on a 10 Mbps hub, each device effectively has ~1 Mbps average.

In a switched network, each port has dedicated bandwidth. With 10 devices on a Gigabit switch, each device has the full 1 Gbps available for communication with any other device (limited by the switch's backplane capacity).

Full-Duplex Operation:

Switches enable full-duplex communication—devices can send and receive simultaneously:

• 1 Gbps full-duplex = 2 Gbps aggregate (1 Gbps each direction)
• Eliminates collisions entirely
• CSMA/CD is disabled in full-duplex mode

Switch Architecture:

Modern switches are sophisticated devices with:

• ASIC-based forwarding: Hardware application-specific integrated circuits process frames at wire speed
• Multi-gigabit backplanes: Internal fabrics capable of handling aggregate traffic from all ports simultaneously
• Port buffers: Memory to handle temporary congestion
• Management processors: CPUs for configuration, monitoring, and control plane operations

Virtual LANs (VLANs) partition a physical LAN into multiple logical networks. Devices in the same VLAN communicate as if on the same physical network, regardless of physical location. Devices in different VLANs require routing to communicate—even if connected to the same switch.

Why VLANs Matter:

IEEE 802.1Q VLAN Tagging:

When frames traverse trunk links (links carrying multiple VLANs), they need a VLAN identifier. IEEE 802.1Q defines a 4-byte tag inserted between the source MAC and EtherType fields:

802.1Q Tag Structure:

With 12 bits for VLAN ID, 802.1Q supports 4094 usable VLANs (0 and 4095 are reserved).

Port Types:

• Access Ports: Connect end devices; frames are untagged. The switch adds/removes tags based on port VLAN assignment.
• Trunk Ports: Connect switches; frames carry 802.1Q tags. A 'native VLAN' handles untagged frames on trunks.
• Hybrid Ports: Carry some VLANs tagged and others untagged (less common in enterprise networks).

Inter-VLAN Routing:

Devices in different VLANs cannot communicate directly—they require routing. Options include:

• Router-on-a-stick: Single router interface with subinterfaces for each VLAN
• Layer 3 switch: Switch with integrated routing capability (SVI - Switch Virtual Interfaces)
• External router: Dedicated routing device with multiple interfaces

Spanning Tree Protocol (STP) prevents loops in switched Ethernet networks. Without loop prevention, a single broadcast frame in a looped topology would circulate forever, consuming all bandwidth—a broadcast storm.

Why Loops Are Deadly:

Consider two switches connected by two links:

1. A broadcast frame arrives at Switch A
2. Switch A floods it out all ports, including both links to Switch B
3. Switch B receives the frame on both links
4. Switch B floods the frame out all ports, including back to Switch A
5. This continues indefinitely, exponentially

Within seconds, the network becomes completely saturated. CPU utilization on switches spikes. Legitimate traffic cannot get through. This is why loops must be prevented.

How STP Works:

STP (IEEE 802.1D) creates a loop-free logical topology by blocking redundant paths:

1. Root Bridge Election: Switches exchange BPDUs (Bridge Protocol Data Units) and elect a root bridge based on lowest Bridge ID (priority + MAC address)

2. Root Port Selection: Each non-root switch selects one root port—the port with the lowest cost path to the root bridge

3. Designated Port Selection: For each network segment, one designated port is selected (the port with lowest cost to root bridges the segment)

4. Port Blocking: All other ports are placed in blocking state—they don't forward data frames but continue listening for BPDUs

STP Convergence:

When topology changes occur (link failure, new switch added), STP must reconverge. Classic 802.1D convergence takes 30-50 seconds—unacceptable for modern networks.

Rapid Spanning Tree (RSTP - 802.1w):

RSTP provides sub-second convergence through:

• Elimination of listening/learning delays in most cases
• Proposal/Agreement mechanism for rapid port transitions
• Edge ports (portfast) that transition immediately
• Merge of blocking and listening into 'discarding' state

Multiple Spanning Tree (MSTP - 802.1s):

MSTP maps multiple VLANs to spanning tree instances. Instead of running 100 separate STP calculations for 100 VLANs, VLANs are grouped into a smaller number of instances. This provides:

• Load balancing across redundant links (different VLANs use different paths)
• Reduced CPU overhead from fewer STP instances
• Interoperability between different vendor implementations

While Ethernet dominates today's LANs, understanding alternative technologies provides historical context and explains why certain design principles exist. Some alternatives also persist in specialized applications.

Token Ring (IEEE 802.5):

IBM's Token Ring (1985) used token-passing for media access:

• Devices arranged in a logical ring
• A 'token' circulates; only the token holder can transmit
• Deterministic access—worst-case latency is predictable
• No collisions by design

Why Token Ring Lost:

• Higher cost per port than Ethernet
• More complex implementation
• Ethernet's speed increases outpaced Token Ring
• First-mover advantage in the market

FDDI (Fiber Distributed Data Interface):

FDDI (1987) operated at 100 Mbps over fiber:

• Dual counter-rotating rings for redundancy
• Token-passing media access
• 100 km maximum ring circumference
• Commonly used as campus backbone before Gigabit Ethernet

ATM (Asynchronous Transfer Mode):

ATM emerged in the 1990s as a unified networking technology for LANs, WANs, and telecommunications:

• Fixed 53-byte cells: 5-byte header + 48-byte payload
• Connection-oriented: Virtual circuits established before data transfer
• Quality of Service: Built-in support for different traffic classes (CBR, VBR, ABR, UBR)
• Cell switching: Hardware-efficient due to fixed cell size

ATM's LANs (LANE): ATM LAN Emulation (LANE) provided Ethernet-like service over ATM infrastructure. However, the complexity and cost couldn't compete with switched Ethernet.

ATM's Legacy: While ATM failed in the LAN, its influence persists:

• QoS concepts influenced DiffServ and IntServ
• Cell-based switching influenced MPLS
• Virtual circuit concepts persist in MPLS-TP and Carrier Ethernet

Wireless LAN (IEEE 802.11):

Wi-Fi deserves special mention as the only significant alternative to Ethernet that succeeded:

• Uses CSMA/CA (Collision Avoidance) for media access
• Operates in ISM bands (2.4 GHz, 5 GHz, 6 GHz)
• Speeds from 2 Mbps (802.11, 1997) to 9.6 Gbps (802.11ax/Wi-Fi 6)
• Complements rather than replaces wired Ethernet

We've explored the technologies that power Local Area Networks, with deep focus on Ethernet—the dominant technology—while understanding the historical alternatives that shaped network design principles.

What's Next:

Now that we understand LAN technologies, we'll explore the components that comprise a LAN. The next page examines switches, access points, cabling, and other infrastructure elements—how they work, how to select them, and how they combine to create functional networks.