Glossary of Terms
A
- Adiabatic cooling
- Cooling air by evaporating water into it, using far less energy than mechanical refrigeration.
- Arc flash
- A dangerous explosive release of energy when electricity jumps through air between conductors.
- ASHRAE
- The American Society of Heating, Refrigerating and Air-Conditioning Engineers; its TC 9.9 committee publishes data-centre thermal guidelines.
- ATS
- Automatic Transfer Switch — moves an electrical load from the utility supply to a generator (and back) automatically.
B
- BESS
- Battery Energy Storage System — large battery installation that stores energy for backup or grid services.
- BMS
- Building Management System — software and controllers that monitor and control mechanical plant such as chillers and air handlers.
- Business continuity
- Keeping critical services running through disruption, using plans, alternative sites and tested recovery procedures.
- Busway
- An overhead enclosed metal bar system that distributes power along a row; racks tap into it with plug-in units.
C
- CDU
- Coolant Distribution Unit — pumps and controls the liquid loop that feeds direct-to-chip cold plates, with a heat exchanger to facility water.
- Chiller
- A machine that removes heat from water using a refrigeration cycle, producing chilled water for cooling.
- Cold aisle
- The aisle where server fronts face each other and draw in cool air.
- Colocation
- Renting space, power and cooling in a data centre owned by a provider, while you own the IT equipment.
- Commissioning
- The structured testing (Levels 1-5) that proves a facility's systems work as designed before going live.
- Concurrent maintainability
- Every component can be taken out for planned maintenance without shutting down IT; the key Tier III requirement.
- CRAC / CRAH
- Computer Room Air Conditioner (has its own compressor) / Computer Room Air Handler (uses chilled water).
- Critical path
- The longest chain of dependent tasks in a project schedule; any delay on it delays the finish date.
- Cross-connect
- A physical cable connecting two customers or a customer and a carrier inside a data centre.
- CUE
- Carbon Usage Effectiveness — total CO2 emissions from data-centre energy divided by IT energy.
D
- DCIM
- Data Centre Infrastructure Management — software that tracks assets, power, cooling and capacity in one view.
- Delta-T
- The temperature difference between incoming and outgoing air or water; higher delta-T usually means more efficient cooling.
- Direct-to-chip cooling
- Liquid cooling where coolant flows through cold plates mounted directly on CPUs and GPUs.
- Disaster recovery (DR)
- Restoring IT systems and data at another site after a serious failure such as fire, flood or cyber attack.
E
- Economiser
- Equipment that uses cool outside air or water to cool the data centre without running compressors (free cooling).
- Edge data centre
- A small facility placed close to users or devices to reduce latency.
- EPMS
- Electrical Power Monitoring System — monitors meters, breakers and power quality across the electrical system.
F
- Fault tolerance
- The ability to keep IT running through any single unplanned failure; the key Tier IV requirement.
G
- Generator
- An engine-driven machine (usually diesel) that produces electricity when the utility supply fails.
- GMP (Guaranteed maximum price)
- A construction contract ceiling price; the contractor pays overruns unless the owner changes the scope.
- Grey space
- Back-of-house areas holding electrical and mechanical plant such as UPS, switchgear and chillers.
H
- Hot aisle
- The aisle where server backs face each other and exhaust hot air.
- Hyperscale
- Very large data centres or campuses run by cloud and internet giants, often hundreds of megawatts.
I
- Immersion cooling
- Submerging servers in a non-conductive fluid that absorbs heat directly (single-phase or two-phase).
- IST
- Integrated Systems Test — Level 5 commissioning, testing all systems together under simulated real conditions.
K
- kW per rack
- Power drawn by one rack; a key measure of density (about 5-15 kW for traditional IT, 100 kW+ for AI racks).
L
- LOTO
- Lockout/Tagout — locking and labelling energy sources so equipment cannot be re-energised during work.
M
- Meet-me room (MMR)
- A secure room where carriers and customers interconnect their networks.
- MOP
- Method of Procedure — a step-by-step approved plan for a specific maintenance or change task.
- MW (megawatt)
- One million watts; data-centre capacity is usually quoted in MW of IT load.
N
- N, N+1, 2N
- Redundancy levels: N = just enough, N+1 = one spare, 2N = a full duplicate system.
- NOC
- Network/Operations Centre — the room where staff monitor the facility and network 24/7.
O
- OCP
- Open Compute Project — a community sharing open hardware designs for servers, racks and data centres.
- Operational technology (OT)
- The control systems, such as BMS, EPMS, PLCs and equipment controllers, that run physical plant like power and cooling.
P
- PDU
- Power Distribution Unit — distributes power to racks; a rack PDU is the power strip inside a rack.
- PPA
- Power Purchase Agreement — a long-term contract to buy electricity, often from a new wind or solar farm.
- Pre-action sprinkler
- A sprinkler system whose pipes stay dry until a detector confirms fire, reducing accidental water damage.
- PUE
- Power Usage Effectiveness — total facility energy divided by IT equipment energy; 1.0 is the theoretical ideal.
R
- Rack
- A standard metal frame (usually 19 inches wide) that holds servers and network gear, measured in rack units (U).
- Rack unit (U)
- The standard height unit in a rack: 1U = 1.75 inches (44.45 mm).
- Raised floor
- A floor of removable tiles on pedestals, creating a void for cables or cool air.
- Rear-door heat exchanger
- A water-cooled coil fitted to the back door of a rack that removes heat as air leaves the servers.
- Redundancy
- Extra components or paths so one failure does not stop the service.
- Remote hands
- A colocation service where on-site staff perform simple tasks on a customer's equipment.
- RPO / RTO
- Recovery Point Objective (how much data you can lose) and Recovery Time Objective (how long you can be down).
- RPP
- Remote Power Panel — a cabinet of breakers near the racks, fed from a PDU, that supplies individual rack circuits.
S
- SCADA
- Supervisory Control and Data Acquisition — industrial control software used to monitor and operate electrical and mechanical plant.
- SLA
- Service Level Agreement — a contract promising availability, power or response times, with penalties if missed.
- Spine-leaf
- A two-layer data-centre network where every leaf switch connects to every spine switch.
- STS
- Static Transfer Switch — switches a load between two power sources electronically in a few milliseconds.
- Switchgear
- Assemblies of breakers and switches that control, protect and isolate electrical circuits.
T
- TCO (Total cost of ownership)
- All costs of buying, running and financing an asset over its life, used to compare build, lease and cloud options.
- TIA-942
- The Telecommunications Industry Association standard for data-centre infrastructure, with Rated-1 to Rated-4 levels.
- Tier
- Uptime Institute's four-level classification (I-IV) of data-centre infrastructure resilience.
- Transformer
- Equipment that changes voltage, e.g. stepping medium voltage down to low voltage for the building.
- Triple-net (NNN) lease
- A lease where the tenant pays rent plus operating costs such as power, taxes, insurance and maintenance.
U
- UPS
- Uninterruptible Power Supply — batteries and electronics that keep IT running instantly when power fails.
- Uptime / nines
- Availability expressed as a percentage, e.g. 99.99% ('four nines') is about 53 minutes of downtime per year.
V
- VESDA
- Very Early Smoke Detection Apparatus — an aspirating system that samples air to detect smoke very early.
W
- White space
- The room where IT racks live, also called the data hall or computer room.
- WUE
- Water Usage Effectiveness — litres of site water used per kWh of IT energy.