Data center terminology and acronyms for new technicians

Data center terminology is mostly shorthand for three things: the equipment that keeps servers powered and cooled, the paperwork that controls who touches it and when, and the hands-on work on the IT floor. This glossary defines the terms, acronyms and abbreviations a new technician hears in the first week on a data center floor or a build.

Each entry also gives a count: how many of the 3,217 live ads on this board named the term on 22 October 2026. Among equipment terms, power leads: generators were named in 708 ads, UPS in 610 and switchgear in 583, against 476 for chillers, the top cooling term, and 367 for fiber, the top IT floor term. The rarest terms are marked rare in job ads, and entries with no count are terms the board does not measure.

The equipment words the ads use most

Generator

708

UPS

610

Switchgear

583

BMS

480

Chiller

476

PDU

352

CRAC

321

Cooling tower

306

EPMS

288

CRAH

245

ATS

205

A count shows how often a word turns up in job descriptions, not whether you need it on day one, and counts move daily. The board leans toward construction (974 of the 3,217 ads), which lifts build-side words. Employers have house acronyms too: when you hear one that is not here, ask.

Power terms, from the utility to the rack

How power reaches a server

Utility

Medium-voltage feed

→

Switchgear

Breakers divide the power

→

ATS

Generator takes over

→

UPS

Batteries bridge the gap

→

PDU and RPP

Circuits out to racks

→

Rack

A and B feeds

• Medium voltage: equipment above 1,000 volts, such as the utility feed and the transformers that step it down. Working on it takes specific qualification. Named in 104 ads.
• Switchgear: assemblies of circuit breakers, switches and protective relays where power enters the building and is divided up. Operating it is normally planned work, done to a MOP. Named in 583 ads.
• Generator: usually a diesel engine-generator that starts automatically when utility power fails and carries the building until utility power returns. Named in 708 ads.
• ATS: automatic transfer switch. It moves the load from utility to generator power when the utility fails, and back when it returns. The acronym is in 205 ads, and "transfer switch" spelled out, which also covers the STS, in 103.
• UPS: uninterruptible power supply. Batteries or a flywheel, plus power electronics, that carry the critical load through a utility failure without a break until the generators take over. Named in 610 ads.
• STS: static transfer switch. It switches a load between two sources, typically two UPS systems, electronically and in a few milliseconds, fast enough that servers keep running. Named in 119 ads.
• PDU: power distribution unit. Either the floor-standing unit that steps voltage down and feeds branch circuits, or the power strip inside a rack that servers plug into. Named in 352 ads.
• RPP: remote power panel. A cabinet of circuit breakers out on the data hall floor, fed from a PDU, that sends circuits to individual racks. Named in 42 ads.
• Busway: prefabricated runs of copper or aluminum bars in a protective housing, usually overhead, with plug-in boxes that feed the racks below. Named in 52 ads.
• A and B feeds: two independent power paths to each rack. Dual-corded servers plug into both, so either path can fail or be maintained without the server losing power.

Cooling terms

How heat leaves the building

Servers

Exhaust into the hot aisle

→

CRAH

Air across a chilled-water coil

→

Chiller

Takes the heat out of the water

→

Cooling tower

Rejects it outside

• CRAC: computer room air conditioner. A unit with its own compressor and refrigerant circuit that cools and recirculates the room air. Refrigerant work on it requires EPA 608 certification. Named in 321 ads.
• CRAH: computer room air handler. Like a CRAC but with no compressor: fans push room air across a chilled-water coil fed from the central plant. People often say CRAC for both. Named in 245 ads.
• Hot aisle and cold aisle: racks stand in rows, fronts facing across a cold aisle of supply air and backs across a hot aisle of exhaust. Containment encloses one aisle so the two airstreams cannot mix. Named in 42 ads.
• Raised floor: a floor of removable tiles on pedestals, with cold air, and at older sites cabling, running underneath. Some newer halls skip it and run power and cabling overhead. Named in 44 ads.
• Chiller: the machine at the heart of a central cooling plant, which chills water (or a water-glycol mix) for the CRAHs and air handlers. Named in 476 ads.
• Cooling tower: rejects heat from water-cooled chillers to the outside air by evaporating part of the water passing through it. Named in 306 ads.
• Liquid cooling: carrying heat away with liquid instead of air, much closer to the chips. It covers direct-to-chip cooling and immersion, where servers sit in tanks of non-conductive fluid. Named in 104 ads.
• CDU: coolant distribution unit. Pumps, a heat exchanger and controls that keep the liquid loop to the racks separate from the building's water loop. Some rack power strips are also sold as CDUs, so read the ad around it. Named in 57 ads.
• Direct-to-chip: liquid cooling through cold plates fixed directly onto the processors, with coolant piped into and out of each server. Named in 30 ads.

Procedures and paperwork

The procedure words the ads use most

Root cause (RCA)

533

Change management

490

SLA

387

SOP

312

MOP

305

Lockout/tagout

233

NFPA 70E

204

Arc flash

170

EOP

162

• MOP: method of procedure. The approved, step-by-step script for one piece of planned work on critical equipment, including how to back out if something goes wrong. Named in 305 ads.
• SOP: standard operating procedure. Written steps for routine, repeated work such as rounds or a generator test. Named in 312 ads.
• EOP: emergency operating procedure. The steps to follow when a specific failure happens, such as a utility outage or a chiller trip. Named in 162 ads.
• Change management (change control): the process every planned change goes through before anyone touches production, from request and review to approval, scheduling and a record afterward, usually with a MOP attached. Named in 490 ads.
• RCA: root cause analysis. The investigation after an incident into why it happened, not just what broke, and how to stop it recurring. The acronym or the phrase "root cause" is in 533 ads.
• SLA: service level agreement. The commitments in a customer contract, such as uptime or how quickly a ticket must be answered. Named in 387 ads.
• PM: preventive maintenance. Scheduled inspection and servicing that keeps equipment from failing, usually issued as work orders from the CMMS.
• Lockout/tagout (LOTO): the procedure, required under OSHA's hazardous energy standard, for isolating equipment from its energy sources and locking and tagging the isolation points so nobody can re-energize it while you work. Named in 233 ads.
• Arc flash: the blast of heat and light from an electrical fault through the air. Equipment labels give the arc flash boundary and the energy or PPE category inside it. Named in 170 ads.
• NFPA 70E: the National Fire Protection Association's Standard for Electrical Safety in the Workplace, which sets the rules for shock and arc flash risk, PPE and who counts as a qualified person. Employers usually provide NFPA 70E training. Named in 204 ads.
• OSHA: the federal Occupational Safety and Health Administration. In ads it usually means OSHA safety rules or the OSHA 10 and OSHA 30 cards, which contractors ask for although OSHA itself does not require them. Named in 592 ads.

The IT floor

• White space: the data hall floor where the IT racks stand. Named in 54 ads.
• Gray space: the areas that support the white space, such as electrical rooms, battery rooms, mechanical plant and generator yards.
• U (rack unit): the height measure for rack equipment. 1U is 1.75 inches, and a common full-height rack is 42U.
• Rack and stack: unboxing servers and network gear, mounting them in racks and cabling them up. Named in 52 ads.
• Fiber: fiber-optic cable. Technicians run, label, clean, test and troubleshoot it, and BICSI's installer and technician credentials cover that cabling work. Named in 367 ads.
• Colocation (colo): a data center that rents space, power and cooling to many customers who bring their own servers, by the rack, cage or suite. Equinix, DataBank and Flexential are colocation operators. Named in 280 ads.
• Cross-connect: a fiber or copper cable linking two parties' equipment inside the building, such as a customer and a network carrier. Named in 54 ads.
• Meet-me room (MMR): the room where carriers and customers interconnect, and where the cross-connects between them are patched. Rare in job ads.
• Smart hands and remote hands: on-site work for customers who are not there, such as rebooting a server or swapping a drive on their instructions. Some providers sell basic remote hands and skilled smart hands as separate tiers. Named in 79 ads.
• Break/fix: reactive repair. Something fails, you find the fault and fix it, as opposed to planned installs and changes. Named in 212 ads.
• RMA: return merchandise authorization. The process for sending a failed part back to the manufacturer under warranty and tracking its replacement. Named in 53 ads.
• IMAC: install, move, add, change, the routine planned work orders on the IT side. Rare in job ads.

Facility systems and software

• Critical environment: the facilities side of a data center, meaning the power, cooling and controls the servers depend on, and the team that runs them. It was in the titles of 78 live ads, mostly at CyrusOne, JLL and Microsoft; critical environment technician covers the role.
• BMS: building management system. The control and monitoring system for mechanical plant: chillers, CRAHs, pumps, temperatures, humidity and alarms. Some sites call it a BAS, for building automation system. Named in 480 ads.
• EPMS: electrical power monitoring system. The electrical counterpart to the BMS, watching switchgear, UPS, generators and breakers, logging events and raising alarms. Named in 288 ads.
• DCIM: data center infrastructure management. Software that tracks what is installed where, and how much space, power and cooling capacity is left. Named in 102 ads.
• CMMS: computerized maintenance management system. The work-order system: it schedules PMs, records repairs and keeps each asset's history. Named in 300 ads.

Building and commissioning

• Hyperscale: a data center campus built at very large scale by or for one of the biggest cloud and AI companies, such as Amazon, Google, Microsoft and Meta, known as hyperscalers. Named in 577 ads.
• Commissioning (Cx): the staged testing that proves a new building's systems were installed correctly and work as designed before it goes live. It usually runs in five levels, from factory testing to the integrated systems test, set out in the commissioning levels guide. The word is in 1,182 ads, a count that also catches decommissioning, the orderly removal of old equipment.
• IST: integrated systems test. The final commissioning stage: power, cooling and controls tested together under load from load banks, which draw power and give off heat the way servers will, and through deliberate failures such as cutting utility power. Named in 151 ads.
• N+1: one more unit than the load needs. If four chillers carry the load, five are installed, so any one can fail or be serviced. Rare in job ads.
• 2N: two complete, independent systems, each able to carry the whole load, such as fully separate A and B power paths. Rare in job ads.
• Tier: the Uptime Institute's Tier Classification System, which rates a site's infrastructure from Tier I (basic capacity) through Tier II (redundant capacity components) and Tier III (concurrently maintainable, so any component can be serviced without shutting down IT) to Tier IV (fault tolerant).
• PUE: power usage effectiveness. The Green Grid defines it as the energy entering the facility divided by the energy the IT equipment uses. A site spending as much on overhead as on IT scores 2.0, and lower is better. Named in 46 ads.

N, N+1 and 2N, for a load that needs four units

Carrying the loadSpare

N

N+1

2N

Uptime Institute's four Tiers

Tier	Uptime calls it
Tier I	Basic capacity
Tier II	Redundant capacity components
Tier III	Concurrently maintainable: any component can be serviced without shutting down IT
Tier IV	Fault tolerant

Newer terms from AI-era builds

AI halls add a few words of their own. AI data center jobs collects the openings at GPU cloud operators such as CoreWeave, AI labs such as xAI, and bitcoin miners converting their campuses, such as IREN in Childress, Texas. Liquid cooling, CDUs and direct-to-chip, above, belong here too.

• GPU: graphics processing unit, the processor AI models are trained and run on. In an AI hall most racks hold GPU servers, and installing, cabling, testing and swapping them is the floor work. Named in 285 ads.
• InfiniBand: a high-bandwidth, low-latency network standard, maintained by the InfiniBand Trade Association, that links GPU servers into clusters working on one job. It runs on its own switches and cables, separate from the site's Ethernet. Rare in job ads.
• Rack density: the power one rack draws, in kilowatts. Racks full of GPUs draw far more than conventional server racks, which is why AI halls are moving to liquid cooling.

Where to go next

These words make more sense on a shift: a day in the life of a data center technician walks through the rounds, tickets and MOPs. On a resume, list only the systems you have worked on; the data center technician resume guide covers the rest. Or see them in use in live data center technician jobs and critical facilities engineer jobs.