Learn

Data center glossary.

Plain-English definitions of the terms you'll hear in job ads, interviews, and training. 30 core terms, grouped by the system they belong to.

Power & electrical

How a facility keeps clean, continuous power flowing to the equipment.

UPS Uninterruptible Power Supply: A battery-backed system that carries the electrical load for a short time when utility power drops, holding equipment up until generators start or power returns.
PDU Power Distribution Unit: Equipment that takes incoming power and splits it out to racks and devices. A floor or rack PDU distributes and, in many cases, meters the power each rack draws.
ATS Automatic Transfer Switch: A switch that automatically moves the load from one power source to another, for example from utility to generator, without someone flipping it by hand.
Generator: An engine-driven backup power source, usually diesel, that supplies the site during an extended utility outage after the UPS bridges the first seconds or minutes.
Busway: An overhead track of enclosed conductors that distributes power along a row of racks, letting technicians tap off power where they need it instead of running individual whips.

Keeping temperature and airflow inside the safe range for the hardware.

CRAC Computer Room Air Conditioner: A room cooling unit that uses a refrigerant compressor to remove heat from the data hall, much like a large, precise air conditioner built for continuous duty.
CRAH Computer Room Air Handler: A room cooling unit that uses chilled water and fans instead of its own compressor. The cooling is produced centrally by chillers and delivered as cold water.
Hot aisle / cold aisle: A rack layout where cold supply air and hot exhaust air are kept on separate aisles so they don't mix, which makes cooling far more efficient.
Containment: Physical barriers (doors, panels, or roofs over an aisle) that seal hot and cold air apart so the cooling system doesn't waste energy mixing them.
Chiller: Plant equipment that produces chilled water for CRAH units and other cooling loops by rejecting heat to the outside, often the largest single energy user on a site.

How equipment is mounted, measured, and organized on the floor.

Rack unit (U): The standard height measurement for rack-mounted equipment. One U is 1.75 inches. A "1U server" takes a single slot; a full rack is commonly 42U or 48U.
Cabinet / rack: The vertical frame or enclosure that servers, switches, and cabling mount into. Cabinets are usually enclosed with doors; open racks are frames without side panels.
Remote hands: On-site technician work performed on behalf of a customer or remote team, swapping a drive, reseating a cable, power-cycling a device, so they don't have to travel to the site.
Blanking panel: A solid filler plate that covers empty rack slots so cold supply air can't short-circuit through gaps to the hot side, protecting airflow separation.

The physical layer that connects everything and keeps it traceable.

Structured cabling: A planned, standardized cabling system with defined pathways, labeling, and connection points, so that connections are consistent, documented, and easy to trace.
Patch panel: A mounted panel of ports that terminates permanent cabling, letting technicians make and change connections with short patch cords instead of re-running fixed cable.
Cross-connect: A managed physical connection between two parties or systems, usually made at a patch panel or meet-me area, used heavily in colocation to link customers and carriers.
Single-mode / multimode fiber: Two fiber-optic cable types. Single-mode carries light over long distances with a narrow core; multimode suits shorter runs inside a facility with a wider core.
MDA / HDA Main / Horizontal Distribution Area: Standard cabling zones in a data center. The MDA is the central connection point; HDAs distribute out toward the equipment rows.

How sites stay running when a component or a whole path fails.

N+1: A redundancy level where there is one spare unit beyond what the load needs. If "N" units are required, the site runs N+1 so any single unit can fail or be serviced without impact.
2N: A redundancy level with two complete, independent systems, full duplication. Either side can carry the entire load, so an entire path can fail without downtime.
Redundancy: Designing spare capacity or duplicate paths into power, cooling, and network so that a single failure doesn't take the load down.
Single point of failure SPOF: Any one component whose failure would bring down a service because nothing backs it up. Reliable designs work to identify and eliminate SPOFs.
Concurrent maintainability: A design property where any single component can be taken offline for maintenance while the site keeps running the load at full function.

The day-to-day process, response, and safety language of the floor.

Runbook: A documented set of steps for handling a specific task or situation, so that any qualified technician can carry it out consistently and safely.
Incident: An unplanned event that disrupts or threatens a service. Data center operations track, escalate, and resolve incidents against defined response targets.
Escalation path: The defined order in which an issue is raised to more senior or specialized people when it can't be resolved at the current level or within a time limit.
SLA Service Level Agreement: A commitment to a measurable level of service, such as uptime or response time, that operations teams are expected to meet and report against.
Change management: The controlled process for planning, approving, and recording changes to live systems so that risky work is reviewed and reversible rather than ad hoc.
LOTO Lockout / Tagout: A safety procedure that isolates energy sources and locks them off before work, preventing equipment from being powered on while someone is working on it.

Ready to practice?

Start with 25 free questions, or get the full Question Bank with role-tagged questions, answer frameworks, and red flags.