How the Switch Processes Frames
We have covered the MAC address and the MAC address table. Now we put them together and follow a single frame from the instant it arrives on a port to the instant it leaves. This is the switch’s core algorithm — learn, look up, forward — and it is also where the switching modes (store-and-forward, cut-through, fragment-free) change what happens.
- List the switch’s processing steps in order, from receiving a frame to aging.
- Explain why a frame with a bad FCS is discarded.
- Describe how the switch chooses to forward, flood, or filter.
- Contrast store-and-forward, cut-through, and fragment-free switching.
- Recap collision domains and broadcast domains.
- Explain why a switch is “transparent” to IP.
The six processing steps
Every frame that enters a switch port goes through the same sequence. Learn these six steps as a pipeline; almost every switch question is a question about one of them.
- Receive the frame. The frame arrives on the ingress port. The switch buffers or inspects it according to its switching mode (see below).
- Check the FCS. The switch recalculates the checksum and compares it with the Frame Check Sequence in the trailer. If the frame is corrupt, it is discarded — it is never forwarded.
- Learn the source MAC. The switch reads the source address and records that MAC is reachable out of the ingress port, adding or updating the MAC address table.
- Look up the destination MAC. The switch searches the table for the destination address.
- Forward, flood, or filter. Known unicast → forward out one port; unknown unicast / broadcast / multicast → flood out all other ports; destination on the ingress port → filter (drop).
- Update the aging timer. The entry for the source MAC has its age reset, so the switch remembers the address as long as the device keeps talking.
Step 2: the FCS check
The FCS (Frame Check Sequence) is a checksum in the Ethernet trailer, computed by the sender over the whole frame. The receiver recomputes it and compares. If the two do not match, the frame was corrupted somewhere on the link — by noise, a bad cable, a duplex mismatch, or a collision. A switch configured for store-and-forward will discard such a frame. The FCS only detects errors; it does not correct them, and the Data Link layer does not retransmit. (Detecting a problem and dropping the frame is usually enough, because TCP above will retransmit if it matters.)
Forward, flood, or filter
Once the destination MAC has been looked up, exactly one of three things happens:
| Destination | Table result | Switch action |
|---|---|---|
| Known unicast | Found, on a different port | Forward out that one port |
| Known unicast | Found, on the ingress port | Filter — drop, do not forward |
| Unknown unicast | Not found | Flood out all other ports |
| Broadcast / multicast | Group address | Flood out all other ports (unless IGMP snooping applies) |
Switching modes
The switch’s frame-forwarding method decides how much of a frame it must read before it starts sending. There are three modes, and the exam tests their trade-offs:
| Mode | What it does | Strength | Weakness |
|---|---|---|---|
| Store-and-forward | Buffers the entire frame, computes and checks the FCS, and forwards only if the frame is valid. | Most reliable; discards corrupt frames, runts, and giants before they spread. | Highest latency, because it must wait for the whole frame. |
| Cut-through | Starts forwarding as soon as it has read the destination MAC (the first 6 bytes after the preamble), without checking the FCS. | Lowest latency, independent of frame size. | Propagates bad frames — runts and corrupt frames are forwarded onward. |
| Fragment-free | Reads the first 64 bytes before forwarding (also called modified cut-through). | Filters out most collision fragments, which occur within the first 64 bytes. | Latency between the other two; it can still forward some errors. |
Latency and what each mode costs
Latency is the delay a frame experiences crossing the switch. The switching mode sets a floor for it:
- Cut-through has the lowest and most fixed latency — it only waits for the destination address, so a small frame and a large frame are treated the same.
- Fragment-free waits for 64 bytes, so its latency is slightly higher but still constant.
- Store-and-forward has latency that grows with frame size, because the whole frame must be received and checked first. A 1518-byte frame is buffered before any of it moves onward.
This is the central trade-off: reliability versus latency. Cut-through buys speed by giving up error checking; store-and-forward buys correctness by waiting. Modern switches are fast enough that store-and-forward is the usual choice, but the concept is a common exam item.
Collision and broadcast domains
Two domain concepts explain why switches replaced hubs:
- A collision domain is the set of devices whose frames can collide because they share a segment. A hub puts every port in one collision domain; every switch port is its own collision domain, because the switch buffers and forwards per port. This per-port separation is called microsegmentation, and it removed the need for CSMA/CD on full-duplex links.
- A broadcast domain is the set of devices that receive each other’s Layer-2 broadcasts. A switch, by default, keeps all its ports in the same broadcast domain — a broadcast is flooded everywhere. Broadcast domains are divided by routers and by VLANs, not by plain switches.
Transparent to IP
A switch is called transparent because it does not alter the Layer-3 content of the traffic passing through it. It reads only the Ethernet header, then forwards the frame unchanged — same source MAC, same destination MAC, same IP packet inside. It does not modify the IP addresses, does not decrement a TTL, and needs no IP configuration to do its job. The hosts and routers on either side behave as if the switch were invisible, which is exactly what “transparent bridging” (IEEE 802.1D) means. This is the crucial contrast with a router, which is very much not transparent: a router rewrites the Layer-2 header at every hop and makes decisions using the Layer-3 addresses.
| Behaviour | Switch (Layer 2) | Router (Layer 3) |
|---|---|---|
| Decision based on | Destination MAC | Destination IP |
| Modifies frame? | No — forwards unchanged | Yes — builds a new frame each hop |
| Breaks collision domains? | Yes (one per port) | Yes |
| Breaks broadcast domains? | No | Yes |
Exam scenario
The question. A PC sends a unicast frame to a server on the same VLAN. Before any other host has spoken, the switch has only one entry in its MAC address table: the server, on Gi0/20. The PC is on Gi0/1. A technician sees the frame leave every port except Gi0/1. Why did the switch flood a frame that was clearly destined for the server?
The answer. Look carefully at the table: the entry the switch has is the server, and the newly arrived frame is from the PC to the server. The switch learns the PC’s source address (adding “PC → Gi0/1”), then looks up the destination. Here is the trap: a switch forwards a frame as a known unicast only if the destination MAC is already in the table. If the server entry was for a different VLAN or had aged out, the destination is unknown, so the switch floods it out of all ports except the ingress port. Once the server replies and the switch has both addresses, later frames are forwarded to the single correct port. This is the classic unknown unicast flooding behaviour, and it is normal and expected — a switch never drops an unknown frame on an access port; it floods it to be safe.
The vocabulary you need
Every term below appears in the lesson text above. Select one to read its definition, then use the quiz at the end to check what stuck.
Review
Check what you have learned with these review questions.
Sources & references
- Cisco. CCNA (200-301) Exam Topics. cisco.com.
- IETF. Request for Comments (RFCs). rfc-editor.org.
- IANA. Internet Assigned Numbers Authority registries. iana.org.
Target exams: CCNA 200-301 v1.1 · CyberOps Associate 200-201. Last content verification: 2026-10-11.