Framing
Framing is how the Data Link layer turns an endless stream of bits into discrete, identifiable units that can be addressed, checked, and delivered. Before you can read the exact Ethernet fields in lesson 08, you need the general shape of a frame and the reasons behind its minimum and maximum sizes.
Why frames exist
The Physical layer has no concept of a “message”. It delivers a continuous stream of bits: 1011010011101… with no pauses and no meaning beyond the signal itself. If the Data Link layer simply handed that stream to the Network layer, the receiver would have no idea where one unit of data ended and the next began.
Framing solves this by grouping bits into a defined unit — the frame — and marking its boundaries with a delimiter. In effect, framing does three things:
- Groups the bits into a unit with a clear start and end.
- Delimits the start and stop, so the receiver can find where each frame begins (the start delimiter) and where it ends.
- Frames the payload with the control fields needed for addressing and error detection.
This is the Data Link layer’s first service (from lesson 05). Without reliable frame boundaries, every field that follows a lost delimiter would be read at the wrong offset and the link would be useless.
What the header carries
The header holds the fields needed to get the frame across the local link and to tell the receiver what to do with it.
- Addressing — a source and destination Layer 2 address (in Ethernet, the 6-byte MAC addresses) so the frame reaches the right device on the link.
- Type / length — a field that says what kind of payload is inside (for example, an IPv4 packet) or, in IEEE 802.3 framing, how long the payload is.
- Control — protocol-specific flags, sequence numbers, or priority fields depending on the technology.
- Start delimiter — in many technologies the header begins with a pattern that marks the frame’s start (Ethernet’s preamble and SFD lean on this idea).
What the trailer carries
The trailer sits after the payload and, on essentially every Layer 2 protocol, holds exactly one thing: the Frame Check Sequence (FCS). The sender computes a CRC over the frame’s contents and writes the result into the FCS; the receiver recomputes it and compares. A mismatch means the frame was corrupted in transit.
Minimum and maximum frame sizes
Ethernet does not allow a frame to be any length. It imposes a floor and a ceiling, each for a different engineering reason.
Why a minimum size?
On a shared half-duplex Ethernet segment, a station detects a collision by listening while it transmits. For detection to be possible, the station must still be transmitting when the collision signal, travelling at the signal propagation speed, returns from the far end of the segment. If the frame were too short, the sender could finish before the collision arrived, miss it, and never back off. So a minimum frame size guarantees the sender is still on the wire long enough to notice a collision. For classic Ethernet at 10 Mbps this works out to 64 bytes. (See lesson 08 for the exact byte math.)
Why a maximum size?
A maximum prevents one station from monopolising the shared medium with an enormous frame, and it bounds the work and memory every device must commit.
- Fairness — no single station can hold the medium for an unbounded time, so other stations still get a turn.
- Bounded buffering — receivers reserve buffer space sized to the maximum frame, so a smaller maximum means cheaper, more predictable memory.
- Error efficiency — a corrupted large frame wastes proportionally more bandwidth when it is discarded, and a single bit error destroys the whole frame; keeping frames bounded limits that loss.
For standard Ethernet the maximum frame is 1518 bytes (1522 with an 802.1Q VLAN tag). Lesson 08 works through both numbers field by field.
MTU: the payload limit
The Maximum Transmission Unit (MTU) is the largest payload a link can carry — it measures the data field, not the whole frame. For standard Ethernet the MTU is 1500 bytes. Confusing MTU with the full frame size is a classic mistake.
| Quantity | What it counts | Standard Ethernet value |
|---|---|---|
| MTU | The payload (data field) only | 1500 bytes |
| Maximum frame | Header + payload + trailer | 1518 bytes |
| Minimum frame | Header + minimum payload + trailer | 64 bytes |
If a packet from Layer 3 is larger than the MTU, it must be fragmented (IPv4) so each piece fits in a frame. A path whose smallest MTU is lower than the packet size is often called a path MTU problem — another lesson learned from framing limits.
Framing on different media
Framing is universal, but the delimiter mechanism adapts to the medium. On a shared LAN (Ethernet), frames are delimited with a preamble and start-frame delimiter and end with an FCS. On a point-to-point link, there is only one possible receiver, so no addressing for a shared medium is needed — the design instead focuses on cleanly marking frame boundaries.
PPP (Point-to-Point Protocol) frames every packet with the same header/payload/trailer skeleton, but uses a flag byte — the value 0x7E (01111110) — at both the start and the end of the frame. If that flag value happens to appear inside the payload, PPP must escape it (byte stuffing for octet-synchronous links, or bit stuffing in HDLC-style links) so the receiver cannot mistake payload data for a delimiter. This is the same principle as Ethernet’s preamble: mark the boundaries so synchronisation never depends on the data content.
Review
Check what you have learned with these review questions.
Sources & references
- IEEE. IEEE 802.3 (Ethernet frame format, preamble/SFD, minimum and maximum frame sizes). standards.ieee.org.
- IEEE. IEEE 802.2 (Logical Link Control framing concepts). standards.ieee.org.
- Simpson, W. (ed.). RFC 1662: PPP in HDLC-like Framing (1994) — flag bytes and byte stuffing. rfc-editor.org/rfc/rfc1662.
- Cisco. CCNA (200-301) Exam Topics, Network Fundamentals. cisco.com.
Target exams: CCNA 200-301 v1.1 · CyberOps Associate 200-201. Last content verification: 2026-10-11.