1. Introduction to Ethernet
Key Concept: Ethernet is the most widely installed local area network (LAN) technology, using a bus or star topology and supporting data transfer rates from 10 Mbps to 400 Gbps.
What is Ethernet?
Ethernet is a family of computer networking technologies commonly used in local area networks (LAN), metropolitan area networks (MAN), and wide area networks (WAN). It was commercially introduced in 1980 and first standardized in 1983 as IEEE 802.3.
Key Characteristics
📡 Shared Medium
Originally used coaxial cable as a shared bus; modern implementations use twisted pair or fiber with switches.
🔀 Frame-Based
Data is transmitted in variable-length frames with source/destination MAC addresses and error checking.
⚡ CSMA/CD
Carrier Sense Multiple Access with Collision Detection - the original access control method for shared Ethernet.
🌐 Scalability
Supports speeds from 10 Mbps (Original) to 400 Gbps (Modern), with backward compatibility.
Why Study Ethernet?
- Ubiquity: Over 85% of wired LANs worldwide use Ethernet
- Foundation: Understanding Ethernet is crucial for TCP/IP, VLANs, and modern networking
- Industry Standard: Essential knowledge for network engineering careers
- Evolution: Teaches fundamental concepts that apply to modern high-speed networks
2. History and Evolution
1973
Xerox PARC Invention
Robert Metcalfe at Xerox PARC invents Ethernet, inspired by ALOHAnet. Initial speed: 2.94 Mbps over coaxial cable. Named after the luminiferous aether once believed to carry light.
1980
DIX Standard
DIGITAL, Intel, and Xerox (DIX) release Ethernet Version 1.0 (10 Mbps). This becomes the de facto standard before IEEE involvement.
1983
IEEE 802.3 Standardization
IEEE publishes 802.3 standard, slightly different from DIX. CSMA/CD becomes the official access method. 10BASE5 (Thick Ethernet) specified.
1985
10BASE2 (Thin Ethernet)
Cheaper, more flexible thin coaxial cable (RG-58) introduced. Easier installation but shorter distances (185m vs 500m).
1990
10BASE-T (Twisted Pair)
Revolutionary shift to unshielded twisted pair (UTP) cabling and star topology with hubs. 100m range per segment.
1995
Fast Ethernet (100BASE-TX)
IEEE 802.3u brings 100 Mbps speeds. Introduces auto-negotiation. Full-duplex operation becomes common.
1998
Gigabit Ethernet (802.3z/ab)
1000 Mbps speeds over fiber (z) and copper (ab). Frame bursting and carrier extension introduced for half-duplex.
2002-2006
10 Gigabit Ethernet
10 Gbps speeds (802.3ae). Operates only in full-duplex (no CSMA/CD). LAN PHY and WAN PHY variants.
2010+
Modern Era
40/100 Gbps (2010), 25/50 Gbps (2016), and 400 Gbps (2017) standards. Focus on data centers and cloud computing.
Important: CSMA/CD was eliminated in Gigabit Ethernet and above when operating in full-duplex mode. Modern switched Ethernet uses dedicated bandwidth per port.
3. IEEE 802.3 Standards Overview
Naming Convention
Ethernet standards use a systematic naming format: [Speed][BASE|BROAD][Medium]
Format Breakdown:
10BASE-T = 10 Mbps, BASEband, Twisted pair
100BASE-TX = 100 Mbps, BASEband, Twisted pair (2 pairs)
1000BASE-SX = 1 Gbps, BASEband, Short wavelength fiber
10GBASE-T = 10 Gbps, BASEband, Twisted pair
Physical Layer Standards Comparison
| Standard |
Speed |
Medium |
Max Distance |
Year |
| 10BASE5 |
10 Mbps |
Thick coax (Ø10mm) |
500m |
1983 |
| 10BASE2 |
10 Mbps |
Thin coax (Ø5mm) |
185m |
1985 |
| 10BASE-T |
10 Mbps |
Cat 3 UTP |
100m |
1990 |
| 100BASE-TX |
100 Mbps |
Cat 5 UTP |
100m |
1995 |
| 100BASE-FX |
100 Mbps |
Multi-mode fiber |
2km |
1995 |
| 1000BASE-SX |
1 Gbps |
MMF (850nm) |
550m |
1998 |
| 1000BASE-LX |
1 Gbps |
SMF (1310nm) |
5km |
1998 |
| 1000BASE-T |
1 Gbps |
Cat 5e/6 UTP |
100m |
1999 |
| 10GBASE-T |
10 Gbps |
Cat 6a/7 |
100m |
2006 |
Data Encoding Methods
Manchester Encoding (10 Mbps)
Transition in middle of bit period. 0 = low-to-high, 1 = high-to-low. Self-clocking but 50% efficiency.
4B/5B + MLT-3 (100BASE-TX)
4 data bits → 5 code bits. MLT-3: three-level encoding (0, +V, -V). 80% efficiency, 125 Mbaud.
8B/10B (Gigabit Fiber)
8 data bits → 10 code bits. DC balance, error detection. Used in 1000BASE-X, PCIe, SATA.
PAM16 + DSQ128 (10GBASE-T)
16-level pulse amplitude modulation. 800M symbols/sec, 128-DSQ mapping. Complex but efficient.
4. Ethernet Frame Structure (IEEE 802.3)
Frame Size: Minimum 64 bytes, Maximum 1518 bytes (without VLAN tag) or 1522 bytes (with VLAN tag). Jumbo frames up to 9000 bytes exist but are non-standard.
Frame Format Visualization
Preamble
7 bytes
SFD
1 byte
Dest MAC
6 bytes
Src MAC
6 bytes
Length/Type
2 bytes
Data & Padding
46-1500 bytes
FCS
4 bytes
Detailed Field Descriptions
1. Preamble (7 bytes)
Pattern: 10101010 repeated 7 times (56 bits). Used for clock synchronization. Allows receiver to lock onto sender's clock before actual frame arrives.
2. Start Frame Delimiter (SFD) - 1 byte
Pattern: 10101011. Marks the end of preamble and beginning of the actual frame. Last two bits (11) indicate start of destination address.
3. Destination MAC Address (6 bytes)
48-bit hardware address of the intended recipient:
- First bit (LSB of first byte): 0 = Unicast, 1 = Multicast
- Second bit: 0 = Global (OUI assigned), 1 = Locally administered
- Example:
00-1B-44-11-3A-B7 (Unicast, Global)
4. Source MAC Address (6 bytes)
48-bit hardware address of the sender. Always a unicast address.
5. Length/Type Field (2 bytes)
Dual-Purpose Field:
• If value ≤ 1500 (0x05DC): Length of data field (IEEE 802.3)
• If value ≥ 1536 (0x0600): Protocol Type (Ethernet II)
Common Type values:
0x0800 = IPv4
0x0806 = ARP
0x86DD = IPv6
0x8100 = VLAN Tag
6. Data Field (46-1500 bytes)
Contains the upper-layer protocol data (usually IP packet). If data is less than 46 bytes, padding is added to meet minimum frame size of 64 bytes.
Minimum Frame Size Rationale: 64 bytes ensures proper collision detection. At 10 Mbps with 2500m max cable length, signal propagation time requires at least 51.2 μs (512 bit times) transmission time to detect collisions.
7. Frame Check Sequence (FCS) - 4 bytes
32-bit CRC (Cyclic Redundancy Check) using polynomial:
G(x) = x³² + x²⁶ + x²³ + x²² + x¹⁶ + x¹² + x¹¹ + x¹⁰ + x⁸ + x⁷ + x⁵ + x⁴ + x² + x + 1
Calculated over all fields except preamble, SFD, and FCS itself.
Inter-Frame Gap (IFG)
12 bytes (96 bit times) of silence between frames. Allows stations to prepare for next frame and prevents continuous transmission.
802.1Q VLAN Tag (Optional)
When present, inserted between Source MAC and Length/Type:
TPID
2 bytes
(0x8100)
TCI
2 bytes
(PCP|CFI|VID)
TCI Breakdown: 3 bits Priority (PCP), 1 bit CFI, 12 bits VLAN ID (0-4095, 0x000 reserved, 0xFFF reserved)
5. CSMA/CD Protocol
Full Name: Carrier Sense Multiple Access with Collision Detection
Purpose: Medium Access Control (MAC) protocol for shared Ethernet buses
How CSMA/CD Works
Station A
Station B
Station C
The Algorithm (1-Persistent CSMA/CD)
- Carrier Sense: Listen to the channel before transmitting
- If idle: Transmit immediately
- If busy: Wait until idle, then transmit
- During transmission: Continue listening (Collision Detection)
- If collision detected:
- Stop transmitting immediately
- Send 32-bit jam signal to ensure all stations detect collision
- Wait random time (Binary Exponential Backoff)
- Retry (up to 16 attempts)
Binary Exponential Backoff Algorithm
After n collisions (n ≤ 10):
Choose random number K from [0, 2ⁿ - 1]
Wait time = K × Slot Time (51.2 μs for 10 Mbps)
For 10 ≤ n ≤ 16: Range fixed at [0, 2¹⁰ - 1]
After 16 collisions: Abort transmission and report error
Collision Domain
A collision domain is a network segment where frames can collide with each other. Key rules:
- Repeaters and hubs extend collision domains
- Switches and routers separate collision domains
- Each switch port creates a separate collision domain
- Modern switched Ethernet eliminates collisions in full-duplex mode
Important Timing: Round-trip propagation delay must be less than transmission time. This limits network diameter:
- 10 Mbps: 2500m max (5-4-3 rule)
- 100 Mbps: 205m max (fast collision detection)
- 1 Gbps (half-duplex): 200m with carrier extension
5-4-3 Rule (10 Mbps Ethernet)
To ensure collisions are detected:
- 5 segments maximum
- 4 repeaters maximum
- 3 populated segments (with stations)
Efficiency Analysis
Maximum throughput occurs at optimal load. Efficiency η can be approximated by:
η = 1 / (1 + 6.44a) where a = propagation time / transmission time
For short frames on long networks, efficiency drops significantly.
6. Ethernet Switching
Hubs vs. Switches
| Feature |
Hub (Layer 1) |
Switch (Layer 2) |
| OSI Layer |
Physical Layer |
Data Link Layer |
| Traffic handling |
Broadcasts to all ports |
Forwards to specific port |
| Collision domains |
Single (shared) |
Separate per port |
| Bandwidth |
Shared among all |
Dedicated per port |
| Intelligence |
None (electrical repeater) |
MAC learning, filtering |
Switch Operations
1. MAC Address Learning
Switches build a MAC address table (CAM table) by examining source MAC addresses of incoming frames.
Process:
- Frame arrives at Port 1 with Source MAC A
- Switch records: MAC A is on Port 1 with timestamp
- If MAC A not already in table, add entry
- If exists, update timestamp (aging refreshes entry)
Aging Time: Default 300 seconds (5 minutes). Entries removed if not refreshed.
2. Forwarding/Filtering Decision
Decision Algorithm:
1. Check destination MAC in CAM table
2. If found and on different port → Forward
3. If found and on same port → Filter (drop)
4. If not found or broadcast/multicast → Flood to all ports except source
3. Loop Prevention (STP)
Redundant links create loops causing broadcast storms. Spanning Tree Protocol (802.1D) prevents this:
- Elects Root Bridge (lowest Bridge ID)
- Calculates shortest path to root
- Blocks redundant ports (alternate paths)
- Convergence time: 30-50 seconds (traditional STP)
Switching Methods
Store-and-Forward
Receives entire frame, checks FCS, then forwards. Highest latency but error-free. Standard for modern switches.
Cut-Through
Forwards after reading destination MAC (first 14 bytes). Low latency but may forward corrupted frames. Rare today.
Fragment-Free
Compromise: reads first 64 bytes (minimum frame size) before forwarding. Prevents runt frame forwarding.
Full-Duplex Ethernet
Eliminates CSMA/CD by providing dedicated channels for transmit and receive:
- Requires point-to-point connection (no hubs)
- Uses two pairs in UTP (10BASE-T, 100BASE-TX) or fiber
- Simultaneous send and receive (double bandwidth)
- No collisions, no distance limitations from CSMA/CD
7. Ethernet Evolution & Speeds
Standard Ethernet
1983 • CSMA/CD • Half/Full Duplex
Fast Ethernet
1995 • 802.3u • Auto-negotiation
Gigabit Ethernet
1998 • 802.3z/ab • Frame bursting
10G & Beyond
2002+ • Full-duplex only • Fiber/Copper
Auto-Negotiation (802.3u Clause 28)
Allows Ethernet devices to automatically select best operational mode:
- Exchanges capabilities via Fast Link Pulses (FLP)
- Selects highest common speed and duplex mode
- Priority: 1000BASE-T full > 1000BASE-T half > 100BASE-TX full > 100BASE-TX half > 10BASE-T full > 10BASE-T half
Duplex Mismatch: When one side is full-duplex and other is half-duplex, severe performance degradation occurs due to excessive collisions. Always ensure matching duplex settings or use auto-negotiation on both ends.
Flow Control (802.3x)
Prevents buffer overflow in full-duplex links using PAUSE frames:
- Receiver sends PAUSE frame to sender
- Contains timer value (0-65535 quanta of 512 bit times)
- Sender stops transmission for specified time
- Receiver can send PAUSE with timer=0 to resume
Energy-Efficient Ethernet (802.3az)
Reduces power consumption during low link utilization:
- Low Power Idle (LPI) mode
- Quick wake-up for data transmission
- Saves 50%+ power during idle periods
- Supported in 100BASE-TX, 1000BASE-T, 10GBASE-T
8. Virtual LANs (VLANs)
Definition: VLAN (IEEE 802.1Q) logically segments a physical network into multiple broadcast domains, improving security, performance, and management.
VLAN Benefits
- Security: Separate sensitive traffic groups
- Cost reduction: No separate physical networks needed
- Performance: Smaller broadcast domains reduce unnecessary traffic
- Flexibility: Move users between VLANs via software configuration
- Management: Easier network administration
802.1Q Tagging
VLAN identification uses 4-byte tag inserted in Ethernet frame:
- TPID (16 bits): Tag Protocol Identifier = 0x8100
- PCP (3 bits): Priority Code Point (802.1p QoS)
- DEI (1 bit): Drop Eligible Indicator (formerly CFI)
- VID (12 bits): VLAN Identifier (1-4094, 0 and 4095 reserved)
Port Types
Access Port
Connects to end devices (PCs, printers). Belongs to single VLAN. Sends untagged frames.
Trunk Port
Inter-switch link. Carries multiple VLANs. Frames tagged with VLAN ID. Native VLAN sent untagged.
Hybrid Port
Can carry both tagged and untagged traffic simultaneously. Used for special configurations.
Native VLAN
VLAN on which untagged frames are received on a trunk port. Default is VLAN 1. Best practice:
- Change Native VLAN from default (security)
- Ensure matching Native VLAN on both ends of trunk
- Never use Native VLAN for user traffic
VLAN Trunking Protocols
- ISL: Cisco proprietary (obsolete)
- 802.1Q: IEEE standard (industry standard)
- IEEE 802.1ad (QinQ): Double tagging for service provider scenarios
Inter-VLAN Routing
Devices in different VLANs cannot communicate directly (Layer 2 isolation). Requires Layer 3 device:
- Router-on-a-stick: Single physical interface with sub-interfaces per VLAN
- Layer 3 Switch: Switch with routing capabilities (SVIs - Switched Virtual Interfaces)
- Physical router: Separate physical interfaces for each VLAN (inefficient)
9. Self-Assessment Quiz
Click on questions to reveal answers and explanations.
Q1: What is the minimum Ethernet frame size and why does it exist?
Answer: 64 bytes (including headers). This ensures that the transmitting station is still transmitting when a collision reaches it from the farthest point of the network (round-trip propagation time). At 10 Mbps, 512 bits (64 bytes) take 51.2 μs to transmit.
Q2: Differentiate between the Length and Type fields in the Ethernet frame.
Answer: Same 2-byte field serves dual purposes: If value ≤ 1500 (0x05DC), it indicates the Length of data field (IEEE 802.3). If value ≥ 1536 (0x0600), it indicates the Type of protocol carried (Ethernet II, e.g., 0x0800 for IPv4).
Q3: Explain why CSMA/CD is not needed in full-duplex switched Ethernet.
Answer: Full-duplex uses separate channels for transmit and receive (point-to-point connection), so there's no shared medium. Each link has dedicated bandwidth, eliminating the possibility of collisions. Therefore, Carrier Sense and Collision Detection mechanisms are unnecessary.
Q4: What is the Binary Exponential Backoff algorithm, and when is it used?
Answer: Used in CSMA/CD after a collision is detected. The station waits a random time before retransmitting: chooses K from [0, 2ⁿ-1] where n is collision count (max 10). Wait time = K × Slot Time. After 16 collisions, transmission aborts.
Q5: How does a switch learn MAC addresses?
Answer: By examining the Source MAC address of every incoming frame. It records the MAC address, ingress port number, and timestamp in the CAM table. If the MAC already exists, it updates the timestamp (refreshes aging timer).
Q6: What is the purpose of the jam signal in CSMA/CD?
Answer: After detecting a collision, stations send a 32-bit jam signal to ensure all stations on the network detect the collision. It prevents stations from falsely thinking the transmission succeeded and ensures consistent collision detection.
Q7: Decode the Ethernet standard name: 1000BASE-LX.
Answer: 1000 = 1000 Mbps (1 Gbps), BASE = Baseband signaling, LX = Long wavelength (fiber optic, 1310 nm laser). Used with single-mode fiber up to 5km.
Q8: What is the difference between a collision domain and a broadcast domain?
Answer: Collision domain: Network segment where frames can collide. Hubs create one shared collision domain; switches separate them (each port is its own). Broadcast domain: Network segment where broadcast frames are forwarded. Switches forward broadcasts; routers and VLANs separate broadcast domains.
Q9: Explain the 802.1Q tag fields and their sizes.
Answer: 4-byte tag inserted after source MAC: TPID (2 bytes, 0x8100), TCI (2 bytes) containing: PCP (3 bits, priority), DEI (1 bit, drop eligible), VID (12 bits, VLAN ID 1-4094).
Q10: Why can't Gigabit Ethernet use CSMA/CD effectively over the same distances as 10 Mbps Ethernet?
Answer: At 1 Gbps, bit time is 1 ns vs 100 ns at 10 Mbps. To maintain CSMA/CD over 2500m would require impossibly large minimum frame sizes (6400 bytes). Solution: Carrier Extension (padding short frames to 512 bytes) or operate only in full-duplex mode (most common).
Q11: What happens when a switch receives a frame with a destination MAC address not in its CAM table?
Answer: The switch floods the frame out all ports except the port on which it was received (unknown unicast flooding). This ensures the destination receives the frame while the switch learns its location from the response.
Q12: Calculate the efficiency of a 10 Mbps Ethernet transmitting 64-byte frames over a 2500m cable (signal propagation speed = 2 × 10⁸ m/s).
Answer: Propagation delay = 2500/(2×10⁸) = 12.5 μs. Round-trip = 25 μs. Transmission time = (64×8)/10⁷ = 51.2 μs. a = 25/51.2 = 0.488. Efficiency ≈ 1/(1+6.44×0.488) ≈ 24%. Short frames are inefficient on long networks!
10. Key Terms Glossary
MAC Address (Media Access Control)
48-bit hardware address burned into network interface cards. First 24 bits (OUI) identify manufacturer, last 24 bits are unique device identifier. Format: XX-XX-XX-XX-XX-XX in hexadecimal.
CSMA/CD (Carrier Sense Multiple Access/Collision Detection)
MAC protocol where stations listen before transmitting, transmit if idle, and stop if collision is detected, using binary exponential backoff for retransmission. Used in shared Ethernet.
Full-Duplex
Communication mode allowing simultaneous transmission and reception on separate channels. Eliminates collisions and doubles bandwidth compared to half-duplex.
Collision Domain
Network segment where data frames can collide with one another. Hubs create single collision domains; switches create separate collision domains per port.
Broadcast Domain
Logical network segment where broadcast frames are forwarded to all nodes. Bounded by routers and VLANs; switches forward broadcasts within a domain.
MTU (Maximum Transmission Unit)
Maximum size of protocol data unit that can be transmitted. Standard Ethernet MTU is 1500 bytes. Jumbo frames support up to 9000 bytes.
FCS (Frame Check Sequence)
32-bit CRC field at end of Ethernet frame used for error detection. Calculated over entire frame except preamble, SFD, and FCS itself.
CAM Table (Content Addressable Memory)
Switch table mapping MAC addresses to physical ports. Hardware-accelerated for fast lookup. Entries age out typically after 300 seconds of inactivity.
VLAN (Virtual LAN)
Logical grouping of network devices that appear to be on the same LAN regardless of physical location. Implemented via 802.1Q tagging to isolate broadcast domains.
Auto-Negotiation
Process where Ethernet devices exchange capabilities and automatically select highest common speed and duplex mode. Uses Fast Link Pulses (FLP).
Inter-Frame Gap (IFG)
96-bit time interval (9.6 μs at 10 Mbps) between Ethernet frames providing recovery time for receivers and preventing continuous transmission.
Baseband vs. Broadband
Baseband uses entire channel bandwidth for single signal (digital). Broadband divides channel into multiple frequency bands for simultaneous analog signals. All modern Ethernet is baseband.
Jam Signal
32-bit signal transmitted by stations upon collision detection to ensure all network participants detect the collision. Prevents partial frame reception.
11. Exam Preparation Checklist
Core Concepts You Should Master:
✓ Frame Structure
- Name and size of each field
- Minimum/maximum frame sizes
- Function of FCS (CRC)
- VLAN tag insertion point
✓ MAC Addresses
- 48-bit format and notation
- OUI (first 3 bytes)
- Unicast vs. Multicast vs. Broadcast
- Locally administered addresses
✓ CSMA/CD
- Algorithm steps
- Binary Exponential Backoff calculation
- Collision detection timing
- Network diameter limitations
✓ Switching
- MAC learning process
- Forwarding vs. flooding vs. filtering
- Collision domain vs. broadcast domain
- STP purpose (basic)
✓ Standards
- Naming convention (10BASE-T, etc.)
- Speed and medium relationships
- Cabling types and distances
- Encoding methods (basic)
✓ VLANs
- 802.1Q tag format
- Access vs. Trunk ports
- Native VLAN concept
- Inter-VLAN routing requirement
Practice Problems:
- Calculate the efficiency of CSMA/CD for given frame sizes and cable lengths
- Determine binary exponential backoff wait times after n collisions
- Identify fields in a given hexadecimal frame dump
- Design network topologies specifying collision and broadcast domains
- Calculate maximum throughput given network parameters
Study Tip: Practice drawing the Ethernet frame format from memory, including field sizes. This is commonly tested in exams. Also, be prepared to explain the difference between switch and hub behavior with specific examples.