Compare physical interface and cabling types

Layer 1 is where networking becomes physical: connectors, wires, glass, and light. The exam expects you to know the copper and fiber cable categories, which Ethernet standard runs over which cable and how far, when a crossover cable is needed, and the difference between shared and point-to-point Ethernet.

1.3.a Single-mode fiber, multimode fiber, copper

Copper: unshielded twisted pair (UTP)

Almost every wired endpoint uses UTP cable: four pairs of copper wires (eight wires total), with each pair twisted around itself. The twisting cancels electromagnetic interference (EMI) and crosstalk (signal leaking from one pair into another). The cable ends in an RJ-45 connector (technically an 8P8C plug). STP (shielded twisted pair) adds a foil or braided shield around the pairs, or around each pair, for electrically noisy environments such as factory floors; it costs more and must be properly grounded. UTP is the default in offices.

Cable quality is rated by category. Higher categories have tighter twists, better insulation, and are certified for higher frequencies:

The standard maximum length for any copper Ethernet run is 100 meters (about 328 feet), including patch cords at both ends.

The Ethernet standards that run over UTP are named by speed, signaling, and medium. “BASE” means baseband (the cable carries one signal), and “T” means twisted pair:

Note the pair count: 10 and 100 Mbps Ethernet use only two of the four pairs (one to transmit, one to receive), while Gigabit and faster use all four pairs in both directions simultaneously. Newer 2.5GBASE-T and 5GBASE-T (802.3bz) run over Cat 5e and Cat 6 respectively and are popular for Wi-Fi 6 access point uplinks.

Fiber optic cable

Fiber carries data as pulses of light through a glass core, surrounded by cladding (glass with a different refractive index that reflects the light back into the core), a protective buffer, and an outer jacket. Fiber is immune to EMI, cannot be tapped easily, carries no electrical current (so it cannot conduct lightning or ground loops between buildings), and spans far greater distances than copper. It costs more and requires more delicate handling and connectors. Fiber is used for switch-to-switch uplinks, between buildings, in data centers, and for WAN circuits.

Two types of fiber exist, differing in core size and light source:

Multimode fiber is graded OM1 through OM5. The wide core lets the light take many paths, and the paths arrive at slightly different times (modal dispersion), which blurs the signal and limits distance. Single-mode’s tiny core allows only one path, so it goes much farther. A common exam distinction: SMF for long distance between buildings or across a city, MMF for inside a building or data center.

The Ethernet fiber standards you should recognize:

“S” means short reach (multimode), “L” means long reach (single-mode), “E” means extended reach.

Fiber connectors include LC (small, the most common today, used on SFP modules), SC (square push-pull), and ST (bayonet, older). Most fiber links use two strands, one to transmit and one to receive, so a patch cable is a duplex pair and the transmit of one end must connect to the receive of the other. Bidirectional (BiDi) optics that use one strand with two wavelengths also exist.

Transceivers: SFP and SFP+

Switches and routers rarely have fiber ports built in. Instead they have empty slots that accept a hot-swappable transceiver module, so the same switch can use short multimode, long single-mode, or even copper on any given port simply by inserting the matching module. The form factors are:

SFP and SFP+ modules are the same physical size; an SFP+ slot usually accepts a 1 Gbps SFP, but an SFP slot cannot run an SFP+ at 10 Gbps. GBIC is the older, larger 1 Gbps form factor. A copper “1000BASE-T SFP” with an RJ-45 jack also exists. Some transceiver pairs must match: the optic on each end of a fiber must use the same wavelength and fiber type. Cisco switches may reject unsupported third-party optics unless service unsupported-transceiver is configured.

Direct-attach cables (DACs) are short (1 to 10 m) copper or active optical cables with SFP+ or QSFP ends permanently attached; they are the cheapest way to connect adjacent switches in a rack at 10 Gbps or more.

1.3.b Connections (Ethernet shared media and point-to-point)

Shared media is how Ethernet began. The original 10BASE5 and 10BASE2 standards ran a single coaxial cable past every computer, and every computer heard every transmission. Even 10BASE-T and 100BASE-T with a hub in the middle behave the same way, because a hub simply repeats every incoming signal out every other port. In a shared-media network only one device can transmit at a time; if two transmit simultaneously, the signals corrupt each other, which is called a collision. All devices sharing the medium form one collision domain.

To manage this, Ethernet uses CSMA/CD (Carrier Sense Multiple Access with Collision Detection):

1. Carrier sense: listen to the wire; if someone is transmitting, wait.
2. Multiple access: when the wire is idle, any device may transmit.
3. Collision detection: while transmitting, keep listening. If the signal on the wire differs from what you sent, a collision occurred. Send a jam signal so everyone discards the damaged frame, wait a random backoff time, and retry.

Because a device cannot transmit and receive at the same time on shared media, it operates in half duplex. The 64-byte minimum frame size exists precisely so that a collision can be detected before the sender finishes transmitting, even on the longest allowed cable.

Point-to-point Ethernet is the modern reality. A switch port connects to exactly one device, and UTP has separate transmit and receive pairs, so both ends can send and receive simultaneously: full duplex. There is no possibility of collision, CSMA/CD is disabled, and each link is its own two-device collision domain. Every 1000BASE-T and faster link is full duplex only; half duplex exists just for 10 and 100 Mbps legacy compatibility. The practical effect is that a full-duplex 1 Gbps link can carry 1 Gbps in each direction at once.

Straight-through versus crossover cables and auto-MDIX

The eight wires in a UTP cable are terminated in one of two color orders defined by the TIA/EIA-568 standard. The T568B order (pin 1 to 8) is white-orange, orange, white-green, blue, white-blue, green, white-brown, brown. T568A swaps the green and orange pairs. A straight-through cable has the same standard on both ends (usually B to B), so pin 1 connects to pin 1, pin 2 to pin 2, and so on. A crossover cable has T568A on one end and T568B on the other, which swaps the orange and green pairs: pin 1 connects to pin 3 and pin 2 connects to pin 6.

Why does this matter? At 10 and 100 Mbps, devices are either MDI (Media Dependent Interface: PCs, routers, servers, which transmit on pins 1 and 2 and receive on 3 and 6) or MDI-X (switches and hubs, which transmit on 3 and 6 and receive on 1 and 2). Connecting an MDI device to an MDI-X device with a straight-through cable lines up transmit with receive naturally. Connecting two devices of the same type requires a crossover so that transmit meets receive.

Modern equipment makes this almost irrelevant. Auto-MDIX (automatic medium-dependent interface crossover) lets a port detect which pins the far end is transmitting on and swap its own internally. Gigabit Ethernet requires all four pairs in both directions anyway and effectively always negotiates the crossover. Auto-MDIX is enabled by default on Cisco Catalyst switches, but it depends on speed and duplex being set to auto; if you hard-code speed or duplex, auto-MDIX may be disabled and the wrong cable type will bring the link down.

SW1(config)# interface gigabitethernet 1/0/10
SW1(config-if)# mdix auto
SW1(config-if)# speed auto
SW1(config-if)# duplex auto