Link Layer and Medium Access
The link layer moves frames between directly connected nodes over a single physical link, and the medium access control sublayer governs how multiple nodes share a broadcast channel without their transmissions destructively colliding.
Definition
The link layer is the protocol layer responsible for transferring frames between two nodes connected by a single link, including framing, link access via a medium access control protocol, and error detection on the link.
Scope
This area covers the second layer of the network stack: framing of data into link-layer frames, error detection and correction over noisy links, and the medium access control (MAC) protocols that coordinate access to a shared communication channel. It includes channel-partitioning, random-access, and taking-turns protocols, switched local-area networks, and the Ethernet family with its switches and addressing. It excludes the physical signaling details below it and the network-layer routing that operates across multiple links above it.
Sub-topics
Core questions
- How is a stream of bits delimited into frames, and how are transmission errors detected or corrected?
- How can many nodes share a single broadcast channel efficiently and fairly?
- What are the trade-offs among channel partitioning, random access, and taking-turns MAC protocols?
- How do learning switches forward frames in a local-area network, and how do they differ from routers?
- Why are link-layer (MAC) addresses needed in addition to network-layer addresses?
Key concepts
- framing
- error detection (parity, checksum, CRC)
- medium access control (MAC)
- channel partitioning (TDMA, FDMA)
- random access (ALOHA, CSMA/CD)
- MAC addresses
- Ethernet
- link-layer switches
- collision and broadcast domains
- address resolution (ARP)
Key theories
- Random multiple access and CSMA/CD
- Random-access protocols let nodes transmit whenever they have data and recover from collisions; carrier-sense multiple access with collision detection (CSMA/CD), used in classic Ethernet, listens before sending and aborts on detecting a collision, then backs off randomly.
- Error detection via redundancy
- Appending check bits computed from the data — parity, checksums, or cyclic redundancy checks — lets a receiver detect (and sometimes correct) bit errors introduced on a link, trading a small overhead for reliability.
- Self-learning switched LANs
- Ethernet switches build forwarding tables automatically by observing the source addresses of incoming frames, forwarding selectively rather than broadcasting, which segments collision domains and scales local-area networks.
Clinical relevance
Link-layer technology is what physically connects devices: Ethernet switches form the backbone of enterprise and data-center networks, Wi-Fi's MAC protocol governs every wireless LAN, and error-detection codes protect data on every link from copper to fiber to radio. Designing low-latency data centers, diagnosing collision and broadcast-storm problems, and segmenting networks with VLANs all rest on link-layer concepts.
History
The ALOHA system at the University of Hawaii (Abramson, early 1970s) pioneered random multiple access over a shared radio channel. Metcalfe and Boggs adapted these ideas into Ethernet at Xerox PARC in 1976, using CSMA/CD over a coaxial cable. Ethernet was later standardized as IEEE 802.3 and evolved from shared coax to switched twisted-pair and fiber, displacing collision-based contention with full-duplex switching.
Key figures
- Robert Metcalfe
- David Boggs
- Norman Abramson
- Andrew S. Tanenbaum
Related topics
Seminal works
- metcalfe1976
- kurose2021
- tanenbaum2010
Frequently asked questions
- What is the difference between a switch and a router?
- A link-layer switch forwards frames within a local-area network based on MAC addresses and is transparent to the network layer, while a router forwards packets between networks based on network-layer (IP) addresses and runs routing protocols. Switches build a single LAN; routers connect different networks together.
- Why do we need MAC addresses if we already have IP addresses?
- MAC addresses identify a network interface on a local link and are used for delivery over that single hop, while IP addresses identify a host within the global addressing structure and are used for end-to-end routing. The two operate at different layers, and address resolution maps an IP address to the MAC address of the next hop.