2.1 Ethernet Framing and Physical Standards

Ethernet Framing and Physical Standards

Ethernet is the ubiquitous foundational technology for Local Area Networks (LANs), Metropolitan Area Networks (MANs), and modern carrier transport backbones. Originally developed in the 1970s at Xerox PARC by Robert Metcalfe and standardized jointly by Digital Equipment Corporation, Intel, and Xerox (the DIX consortium), Ethernet has evolved from a 2.94 Mbps shared coaxial bus into a high-performance, full-duplex switched architecture scaling up to 400 Gbps and 800 Gbps optical transport. Understanding frame encapsulation, bit-level addressing conventions, and transmission modes is essential for service provider engineering on the Nokia 7750 Service Router (SR) platform.

---

Historical Evolution of Ethernet Physical Standards

The physical media and signaling techniques of Ethernet have progressed through several major milestones, gradually moving from shared collision-prone topologies to dedicated point-to-point switched links:

• 10BASE5 (Thicknet): IEEE 802.3 standard operating at 10 Mbps over thick 50-ohm coaxial cable up to 500 meters in length. Required vampire taps and external Attachment Unit Interfaces (AUIs).
• 10BASE2 (Thinnet): 10 Mbps over flexible RG-58 coaxial cable with BNC T-connectors up to 185 meters. While less costly, any physical break in the bus disrupted the entire network segment.
• 10BASE-T: Introduced star topology over Category 3 unshielded twisted pair (UTP) copper cabling with RJ-45 modular connectors up to 100 meters, connecting endpoints to a central multiport repeater (hub).
• 100BASE-TX (Fast Ethernet): IEEE 802.3u standard delivering 100 Mbps over Category 5 UTP using 4B5B block coding and MLT-3 signaling.
• 1000BASE-T / 1000BASE-SX / 1000BASE-LX (Gigabit Ethernet): IEEE 802.3z (fiber) and IEEE 802.3ab (copper). 1000BASE-T uses all four pairs of Cat5e/Cat6 cable simultaneously in full duplex. 1000BASE-SX operates over multimode fiber (850 nm) for short intra-datacenter runs, while 1000BASE-LX uses single-mode fiber (1310 nm) for standard single-mode reaches up to 5 km; extended-reach optics require separate specifications.
• High-speed Ethernet: Encoding, error correction, optics, fiber type, and reach depend on the selected PHY. Do not assume all speeds use one line code.

---

Frame Formats: Ethernet II (DIX) vs. IEEE 802.3

Two primary Ethernet framing formats exist in modern networks: Ethernet II (also called DIX Ethernet) and IEEE 802.3. While both formats share identical preamble, addressing, and checksum fields, they diverge fundamentally in how they identify the payload carried within the frame.

The EtherType vs. Length Demarcation Threshold

To ensure interoperability on the same physical wire, the IEEE established a clean numerical boundary for the 2-byte field following the Source MAC address:

$$\text{Value} \le 1500 \; (0x05DC) \implies \text{IEEE 802.3 Length Field}$$ $$\text{Value} \ge 1536 \; (0x0600) \implies \text{Ethernet II (DIX) EtherType Field}$$

Values between 1501 and 1535 (inclusive) are undefined and treated as illegal/corrupt frames. Ethernet II uses the EtherType field to directly multiplex upper-layer protocols:

• 0x0800 — Internet Protocol version 4 (IPv4)
• 0x86DD — Internet Protocol version 6 (IPv6)
• 0x0806 — Address Resolution Protocol (ARP)
• 0x8100 — IEEE 802.1Q VLAN Tagged Frame
• 0x8847 — MPLS Unicast Label Stack
• 0x8848 — MPLS Multicast Label Stack
• 0x88A8 — IEEE 802.1ad Provider Bridging (QinQ outer Service-VLAN tag)

IEEE 802.3 LLC and SNAP Encapsulation

When the Type/Length field contains a value $\le 1500$, the payload begins with an IEEE 802.2 Logical Link Control (LLC) header. Standard LLC headers include:

• DSAP (Destination Service Access Point): 1 byte identifying the receiving upper-layer process.
• SSAP (Source Service Access Point): 1 byte identifying the sending process.
• Control Field: 1 or 2 bytes indicating unnumbered or connection-oriented service (0x03 = Unnumbered Information).

Because 1-byte SAP fields offered limited protocol identifiers, the IEEE introduced the Subnetwork Access Protocol (SNAP) extension. When DSAP and SSAP are both set to 0xAA and Control is set to 0x03, an 802.2 SNAP header follows. The SNAP header consists of a 3-byte OUI (set to 0x000000 for standard protocols) and a 2-byte Protocol ID, which duplicates standard Ethernet II EtherType numbers. Today, IP traffic across Ethernet networks overwhelmingly utilizes native Ethernet II encapsulation due to its reduced overhead and direct hardware parsing efficiency.

Frame Size Boundaries, MTU, and Jumbo Frames

Ethernet specifications define strict upper and lower limits on frame sizes. Understanding these limits is critical when configuring Maximum Transmission Units (MTUs) across Nokia Service Routers.

The 64-Byte Minimum Frame Size and Slot Time

An untagged standard Ethernet frame has an absolute minimum legal size of 64 bytes (512 bits):

$$\text{Minimum Frame} = 6 \text{ (DMAC)} + 6 \text{ (SMAC)} + 2 \text{ (Type)} + 46 \text{ (Min Payload)} + 4 \text{ (FCS)} = 64 \text{ Bytes}$$

If an application generates an IP packet smaller than 46 bytes (for example, a 40-byte TCP ACK packet without payload: 20 bytes IPv4 header + 20 bytes TCP header), the Ethernet network interface card (NIC) must append 6 bytes of padding (0x00) to satisfy the 46-byte payload requirement.

Physical Layer Overhead Exclusions

The physical layer elements—the 7-byte Preamble, the 1-byte SFD, and the 12-byte Inter-Packet Gap (IPG) (also called Inter-Frame Gap / IFG)—are hardware synchronization tools. They are never counted in the 64-byte minimum or 1518-byte maximum frame size calculations:

• Preamble & SFD (8 bytes): Allow receiving PHY transceivers to lock their phase-locked loops (PLLs) to the incoming bit clock.
• Inter-Packet Gap (12 bytes / 96 bit times): Provides a mandatory recovery delay between consecutive frame transmissions, allowing receiver circuitry to reset.

Maximum Frame Sizes, Tagged Frames, and Jumbo Frames

• Standard Untagged Frame (1518 Bytes): Calculated as 1500 bytes maximum Layer 3 payload (the standard IP MTU) + 14 bytes Layer 2 header (6 DMAC + 6 SMAC + 2 EtherType) + 4 bytes FCS.
• IEEE 802.1Q Tagged Frame (1522 Bytes): Insertion of a 4-byte 802.1Q VLAN header expands the maximum frame size to 1522 bytes without reducing the 1500-byte IP payload.
• QinQ / IEEE 802.1ad Double-Tagged Frame (1526 Bytes): Insertion of two 4-byte VLAN tags (outer Service-VLAN tag and inner Customer-VLAN tag) expands maximum size to 1526 bytes.
• Baby Giant Frames: Frames between 1519 and ~1600 bytes that carry VLAN tags, MPLS shim labels, or PPPoE encapsulation without exceeding typical switch buffer thresholds.
• Jumbo frames: Limits are implementation-specific. A value such as 9000 may describe IP MTU while a port limit counts different bytes. Check port, service, and tunnel definitions rather than treating 9216 as a universal maximum.

---

48-Bit MAC Addressing Architecture (EUI-48)

An Ethernet interface uses a link-layer Media Access Control (MAC) address, formally specified by the IEEE as an EUI-48 (Extended Unique Identifier 48-bit) address.

+---------------------------+---------------------------+
|  24 bits: OUI (Vendor)    |  24 bits: NIC Extension   |
+---------------------------+---------------------------+
| Byte 0 | Byte 1 | Byte 2  | Byte 3 | Byte 4 | Byte 5  |
+--------+--------+---------+--------+--------+---------+
    |       
    +---> Bit 0 (b0, LSB): Individual (0) / Group (1)
    +---> Bit 1 (b1):      Universal (0) / Local (1)

A MAC address contains 48 bits (6 bytes), commonly written as twelve hexadecimal characters. In Nokia SR OS, MAC addresses are formatted in colon-delimited pairs (00:16:4d:11:22:33), whereas other platforms use hyphens (00-16-4D-11-22-33) or dotted-quads (0016.4d11.2233).

OUI and Vendor Assigned Components

1. Organizationally Unique Identifier (OUI) (First 24 bits / 3 bytes): Assigned by the IEEE Registration Authority to hardware manufacturers for the applicable allocation; the hexadecimal values used here are illustrative.
2. Network Interface Identifier (Last 24 bits / 3 bytes): Assigned uniquely by the manufacturing vendor to each physical network port.

The Control Bits: Individual/Group and Universal/Local

The two least significant bits of the first transmitted octet (Byte 0) serve crucial architectural control functions:

• Individual / Group (I/G) Bit (Bit 0 / LSB of Octet 0):
  • 0 = Individual (Unicast): Destination address corresponds to a single, unique network interface.
  • 1 = Group (Multicast / Broadcast): Destination address corresponds to a multicast group or broadcast destination.
• Universal / Local (U/L) Bit (Bit 1 of Octet 0):
  • 0 = Universally Administered: The address uses the universal administration convention; operational uniqueness must still be maintained.
  • 1 = Locally Administered: The address has been overridden or manually assigned by an administrator, hypervisor (e.g., VMware MAC pools), or other local assignment. A virtual address need not have U/L = 1; the standardized IPv4 VRRP address begins with 00.

Address Types and Special MAC Ranges

---

Transmission Modes: Half-Duplex CSMA/CD vs. Full-Duplex

Ethernet interfaces operate in one of two physical transmission modes:

Half-Duplex Operation and CSMA/CD

In half-duplex mode, an interface shares a single physical communication channel for both transmitting and receiving data. Because simultaneous transmission causes signals to collide and corrupt data, stations must adhere to the Carrier Sense Multiple Access with Collision Detection (CSMA/CD) protocol:

1. Carrier Sense: Before transmitting, the station monitors the medium to check for an existing electrical carrier signal. If the wire is busy, it defers transmission.
2. Multiple Access: When the wire is detected idle for at least the Inter-Packet Gap (96 bit times), any station with data ready may begin transmitting.
3. Collision Detection: While transmitting, the station continuously monitors its receiver circuitry. If two stations transmit simultaneously, their signals collide, resulting in an abnormal voltage level or differential signal pattern.
4. Jam Signal: Upon detecting a collision, the station immediately aborts its frame and transmits a 32-bit jam signal. This ensures that all other transmitting stations detect the collision event.
5. Truncated Binary Exponential Backoff: Each station calculates a random backoff delay before attempting retransmission:

$$r \in [0, 2^k - 1] \quad \text{where } k = \min(n, 10)$$

Here, $n$ represents the collision attempt number. The station waits $r \times \text{slot time}$ before sensing the carrier again. If an interface encounters 16 consecutive collisions for a single frame, the frame is aborted and dropped.

Full-Duplex Point-to-Point Operation

Full-duplex Ethernet eliminates shared media entirely. Using simultaneous transmit and receive channels, full duplex provides the following properties. Gigabit copper can use the same pairs in both directions with echo cancellation; separate pairs or fibers are not a universal requirement.

• Simultaneous Bidirectional Data Flow: Stations transmit and receive data concurrently at full line rate without contention.
• Zero Collisions: CSMA/CD contention does not apply to a point-to-point full-duplex link.
• CSMA/CD is Deactivated: Loopback sensing, collision detection circuitry, jam signals, and backoff algorithms are completely disabled.
• Doubled Aggregate Bandwidth: A 10 Gbps full-duplex link delivers 10 Gbps ingress and 10 Gbps egress concurrently (20 Gbps aggregate throughput).

Primary technical reference

Specification or Nokia manual, checked October 20, 2026. Examples are teaching scenarios, not captured router execution.