Overview
Every device understands a specific OSI layer — hub (L1), switch (L2), router (L3). This guide covers device roles, copper and fiber cabling, console access, duplex, and how lightweight APs fit enterprise designs.
Network devices — roles at a glance
Every device on a network operates at one or more OSI layers. Knowing what each device understands tells you where it can break and what it can fix.
| Device | Primary layer | What it understands | Key behavior |
|---|---|---|---|
| Hub (legacy) | L1 Physical | Electrical signals only | Floods all ports — single collision domain |
| Switch | L2 Data Link | MAC addresses, VLANs | Forwards unicast by MAC table; each port = collision domain |
| Router | L3 Network | IP addresses, routing tables | Connects subnets; does not forward broadcasts |
| L3 switch | L2 + L3 | MAC + IP (SVIs) | Switching at wire speed with routed VLAN interfaces |
| Wireless AP | L1 + L2 | 802.11 frames | Bridge wireless clients to wired Ethernet (shared medium) |
| Firewall | L3–L7 | IP, ports, application state | Policy enforcement between security zones |
A hub uses a star physical topology (central device, spokes) but a bus logical topology (all devices hear all frames). Switches maintain star physical and improve logical segmentation via MAC learning.
Network architecture models (Exam 1.2)
CCNA topic 1.2 also tests how you scale beyond a single switch — campus tiers, data center fabric, SOHO, and cloud placement. These are design blueprints, not cable shapes.
| Model | Layers / shape | Best for |
|---|---|---|
| 3-tier campus | Access → Distribution → Core | Large enterprise LAN |
| 2-tier collapsed core | Access + merged core/distribution | Medium campus, lower cost |
| Spine-leaf | Leaf ↔ Spine only (no leaf-to-leaf) | Data center east-west traffic |
| SOHO | Single integrated gateway | Home / small office |
| On-prem / public / hybrid cloud | Where workloads run | CapEx vs OpEx, compliance, elasticity |
North-south = traffic in/out of the data center (users, Internet). East-west = server-to-server inside the DC. Modern DCs are dominated by east-west flows — that is why spine-leaf + ECMP replaced 3-tier in data centers, while 3-tier still fits large campus LANs.
Layer mapping on 3-tier campus (exam favorite):
- Access — PoE, VLAN assignment, port security, end devices
- Distribution — inter-VLAN routing, ACLs between departments, redundant uplinks
- Core — high-speed switching only; avoid heavy ACL inspection here
Spine-leaf rules: every leaf uplinks to every spine; leaves never connect to each other; spines never connect to each other; fixed two-hop path (Leaf → Spine → Leaf) with ECMP instead of STP blocking.
See the full reference with diagrams on Network Topologies & Architecture (includes The CyberSec Guru — 2-tier, 3-tier, spine-leaf as further reading).
Hubs — deprecated but exam-relevant
Hubs are Layer 1 multiport repeaters. They amplify and repeat electrical signals but do not read MAC addresses.
When Device A sends a frame to Device C through a hub:
- Frame arrives at the hub on one port
- Hub floods out all other ports (no intelligence)
- Devices B and D read the destination MAC, see it is not for them, and drop the frame
- Device C accepts and processes the frame
Hub characteristics:
- Single collision domain — CSMA/CD required (half duplex)
- Single broadcast domain — all broadcasts reach every device
- Shared bandwidth — 10 Mbps total across all ports (~30% usable due to collisions)
- Replaced by switches in all modern networks
Switches — Layer 2 forwarding
Switches are bridges implemented in hardware (ASICs) for line-rate forwarding. Each port is its own collision domain, but all VLAN-unsegmented ports share one broadcast domain.
Switch learning process:
- Frame arrives on port 1 from Host A
- Switch records A's MAC → port 1 in the MAC address table
- If destination MAC is unknown, switch floods to all ports except ingress
- When C replies, switch learns C's MAC → port 3
- Future A↔C traffic is forwarded only between ports 1 and 3
Switches give each port dedicated bandwidth (e.g., 1 Gbps per port). Full duplex allows simultaneous send and receive — collision detection is disabled because collisions cannot occur. Mismatched duplex (one side full, other half) causes late collisions and poor performance — always verify both sides.
Routers — Layer 3 connectivity
Routers connect different IP subnets. They strip Layer 2 frames, examine Layer 3 destination addresses, consult the routing table, and re-encapsulate toward the next hop.
Router behaviors you must know:
- Do not forward broadcasts (ARP requests stay local)
- Rewrite MAC addresses at each hop; preserve IP addresses end-to-end
- Require an IP address on each connected interface (
ip address+no shutdown) - Build routing tables from connected interfaces, static routes, and dynamic protocols
Same-subnet communication: Host ARPs for the target directly.
Remote-subnet communication: Host ARPs for its default gateway, sends the frame to the gateway MAC, but the IP header still contains the remote destination.
Wireless access points
Access points operate similarly to hubs on the wireless side — the RF medium is shared. Multiple clients contend for airtime. On the wired side, the AP is typically a bridge into the switched network. Enterprise deployments use lightweight APs managed by a WLC — see the Wireless Architectures guide.
Firewalls — policy enforcement
Firewalls inspect traffic between zones (inside, outside, DMZ). At CCNA scope, know they filter by IP, protocol, and port. Stateful firewalls track connection state. Place ACLs and firewalls where policy requires — not every small network runs a dedicated firewall appliance.
Copper cabling — UTP and standards
Modern Ethernet uses twisted pair copper (UTP or STP):
- UTP — Unshielded Twisted Pair; standard in offices
- STP — Shielded Twisted Pair; noisy environments with extra shielding
- Max distance — 100 meters per segment for copper Ethernet
- Connectors — RJ-45 (8P8C)
10BASE-T (twisted pair) replaced coaxial 10BASE-2 (thinnet) and 10BASE-5 (thicknet). Coax used bus topology with terminators at each end; a single cable break took down the entire segment.
T568A vs T568B pinouts
Both wiring standards produce a straight-through cable when the same standard is used on both ends (pin 1 to pin 1, pin 2 to pin 2, etc.).
| Pin | T568A | T568B |
|---|---|---|
| 1 | White/Green | White/Orange |
| 2 | Green | Orange |
| 3 | White/Orange | White/Green |
| 4 | Blue | Blue |
| 5 | White/Blue | White/Blue |
| 6 | Orange | Green |
| 7 | White/Brown | White/Brown |
| 8 | Brown | Brown |
Gigabit (1000BASE-T) uses all four pairs (all 8 wires). Fast Ethernet used only two pairs for data.
Straight-through vs crossover cables
MDI (Media Dependent Interface) — typical on PCs, routers, servers. MDIX (MDI crossover) — typical on switch ports.
| Connection | Historical cable | Modern (Auto-MDIX) |
|---|---|---|
| PC → Switch | Straight-through | Straight-through (auto negotiates) |
| Switch → Switch | Crossover | Straight-through (auto negotiates) |
| PC → PC | Crossover | Straight-through (auto negotiates) |
| Router → Router | Crossover | Straight-through (auto negotiates) |
Crossover cable swaps transmit and receive pairs so TX on one end connects to RX on the other:
- Pin 1 (TX+) ↔ Pin 3 (RX+)
- Pin 2 (TX−) ↔ Pin 6 (RX−)
Auto-MDIX (1998+) detects the connected device and internally crosses pairs as needed. Crossover cables are rarely needed in modern labs, but CCNA still tests the concept.
Know when crossover was required historically even if your lab switches auto-negotiate. "Like device to like device" = crossover in the pre-Auto-MDIX era.
Fiber cabling
Fiber uses light instead of electrical signals. Two main types:
| Type | Core size | Light path | Distance | Use case |
|---|---|---|---|---|
| Multi-mode (MM) | Larger core (50/62.5 µm) | Multiple light paths | Shorter (hundreds of meters) | Campus, data center |
| Single-mode (SM) | Smaller core (~9 µm) | Single light path | Longer (kilometers) | WAN, long campus runs |
Fiber connectors include LC, SC, and ST (legacy). Fiber is immune to EMI and supports much longer distances than copper.
Console access — out-of-band management
Console access is the first connection to a new Cisco device. It provides direct terminal access independent of network connectivity.
Traditional method:
- RJ-45 console port on the Cisco device
- Rollover cable (console cable) to a serial port
- USB-to-serial adapter on modern PCs
- Terminal emulator: 9600 baud, 8 data bits, no parity, 1 stop bit, no flow control
Modern method:
- USB console port (driver required)
- USB Type-A to USB Mini-B or Micro-B directly to the console port
! Terminal settings Speed: 9600 Data bits: 8 Parity: None Stop bits: 1 Flow control: None
! First boot may show System Configuration Dialog ! Type "no" to enter CLI manually
Once connected, you configure hostname, management IP, and SSH for in-band management going forward.
CSMA/CD — legacy collision handling
On half-duplex shared segments (hubs, old half-duplex switch ports):
- CS (Carrier Sense) — listen before transmitting
- MA (Multiple Access) — any device may transmit when quiet
- CD (Collision Detection) — if collision detected, send jam signal, wait random backoff, retry
Collisions increase with more devices and longer cable runs. Full-duplex switched ports eliminate collisions entirely.
Device selection — design thinking
| Need | Device |
|---|---|
| Connect PCs in one VLAN | L2 switch |
| Connect two subnets | Router or L3 switch SVI |
| Wireless client access | Access point (+ WLC in enterprise) |
| Internet edge policy | Router with ACLs or dedicated firewall |
| Long-distance building link | Single-mode fiber between switches |
Verification commands
show version show cdp neighbors detail show lldp neighbors detail show ip interface brief show interfaces status
CDP is Cisco proprietary, enabled by default, works at L2 without IP. LLDP is the open standard — may be disabled by default on some platforms. Both help map physical connections during troubleshooting.
Exam checklist
- Compare hub, switch, and router at OSI layers
- Map 3-tier campus functions to access, distribution, and core layers
- Contrast spine-leaf (east-west, ECMP) with 3-tier campus (north-south)
- Distinguish SOHO, on-prem, public cloud, and hybrid cloud scenarios
- Explain MAC learning and flooding on a switch
- Describe why routers do not forward broadcasts
- Choose straight-through vs crossover (historical and modern)
- Distinguish multi-mode and single-mode fiber
- Connect to a device via console with correct terminal settings
- Explain full vs half duplex and collision domain boundaries