A–C
ASHRAE
American Society of Heating, Refrigerating and Air-Conditioning Engineers. Publishes thermal guidelines for data center equipment (ASHRAE TC 9.9). ASHRAE A1 class equipment operates at 15–32°C inlet temperature; A3/A4 class supports higher temperatures for improved cooling efficiency.
ATS (Automatic Transfer Switch)
A device that automatically switches between power sources (utility feeds, or utility and generator) when a failure is detected. Transfer time is typically 100–300 milliseconds. Static transfer switches (STS) provide sub-cycle transfer (less than 4ms) for sensitive loads.
Availability
The percentage of time a system or facility is operational. Expressed as a percentage (e.g., 99.982%) or in annual downtime hours. Tier III targets 99.982% (1.6 hours/year); Tier IV targets 99.995% (26 minutes/year).
Busway
An overhead power distribution system consisting of enclosed conductors (bus bars) with tap-off boxes for connecting rack PDUs. More flexible than conduit-based distribution for reconfiguration. Common in modern data centers for high-density power distribution.
Cabinet / Rack
A standardized enclosure for mounting IT equipment. Standard rack width is 19 inches; height is measured in rack units (U), where 1U = 1.75 inches. Standard rack heights: 42U (most common), 45U, 48U. Full-height cabinets are typically 7 feet tall.
Cage
A physically secured area within a colocation facility, enclosed by mesh fencing or solid walls, that provides a dedicated space for a single customer. Cages provide physical separation and access control within a shared facility.
Chiller
A refrigeration system that produces chilled water for cooling data center air handlers (CRAH units). Water-cooled chillers are more efficient than air-cooled but require cooling towers. Chiller COP (Coefficient of Performance) typically ranges from 3.0 to 6.0.
Cold Aisle
The aisle in front of server racks where cold supply air is delivered. Servers draw cold air from the cold aisle and exhaust hot air into the hot aisle behind them. Hot/cold aisle separation is fundamental to efficient data center cooling.
Commissioning
The process of verifying that a data center facility performs as designed. Includes component testing, subsystem testing, integrated systems testing, and operational testing. Commissioning is as important as design — installation errors and integration failures are discovered during commissioning, not during design.
CRAC (Computer Room Air Conditioning)
A self-contained cooling unit with an integral compressor that cools, dehumidifies, and filters air using a direct expansion (DX) refrigeration cycle. No external chiller required. Common in smaller data centers and server rooms.
CRAH (Computer Room Air Handler)
A fan-coil unit that uses chilled water from a central chiller plant to cool air. No integral compressor. More efficient than CRAC at scale. Standard for large data centers with central chiller plants.
Cross-Connect
A physical cable connection between two customers within the same colocation facility, or between a customer and a carrier. Enables direct, low-latency connectivity without traversing the public internet.
D–F
DCIM (Data Center Infrastructure Management)
Software that provides real-time visibility and management of data center infrastructure — power, cooling, space, and connectivity. Enables proactive capacity management, energy optimization, and change management.
DLC (Direct Liquid Cooling)
A cooling approach that brings coolant directly to heat-generating components via cold plates. Significantly more efficient than air cooling for high-density workloads. Required for AI servers above 40 kW per rack and mandatory for NVIDIA GB200 NVL72 systems.
Downtime
The period during which a system or facility is unavailable. Planned downtime occurs during scheduled maintenance; unplanned downtime results from failures. Tier ratings are defined by maximum annual downtime: Tier III allows 1.6 hours/year; Tier IV allows 26 minutes/year.
Dual-Corded
IT equipment with two independent power supplies, each connected to a different power path (A-side and B-side). Essential for Tier III/IV facilities — a failure in either power path does not affect equipment operation because the other path continues to supply power.
Economizer / Free Cooling
A cooling mode that uses outdoor air or water to cool the data center when ambient temperatures are low enough, bypassing or supplementing mechanical refrigeration. Reduces cooling energy consumption significantly in cooler climates.
Fault Tolerance
The ability to continue operating without interruption following any single failure. Tier IV data centers are fault tolerant — any single component failure cannot affect IT operations. Distinct from concurrent maintainability (Tier III), which allows maintenance without downtime but does not protect against unplanned failures.
Floor Loading
The weight-bearing capacity of the data center floor, measured in pounds per square foot (psf) or kilograms per square meter. Traditional IT equipment averages 100–150 psf. AI-dense deployments with liquid cooling can exceed 300 psf. Structural design must account for concentrated loads from high-density racks.
G–I
Generator
A diesel-powered backup power system that provides extended power during utility outages. Must start and reach full load within 10–15 seconds. Sized for 125% of maximum IT load. On-site fuel storage provides 24–72 hours of runtime at full load.
Hot Aisle
The aisle behind server racks where hot exhaust air exits. Hot aisle containment captures this air and routes it back to cooling units, preventing it from mixing with cold supply air.
Hot Aisle Containment (HAC)
A cooling optimization that encloses the hot aisle, capturing hot exhaust air and directing it back to cooling units. More efficient than cold aisle containment because it captures heat at the source.
Hyperscale
Data centers of extreme scale, typically 100,000+ square feet and consuming 100+ MW of power. Operated by cloud providers (AWS, Azure, GCP) and large internet companies. Designed for extreme density and operational efficiency.
Immersion Cooling
A cooling approach that submerges servers directly in dielectric fluid. Achieves 100+ kW per tank with PUE as low as 1.03. Requires modified servers (fans removed) and purpose-built tanks. Increasingly adopted for AI training clusters.
In-Row Cooling
Cooling units installed within server rack rows, directly adjacent to high-density racks. Cools air at the point of heat generation. Effective for rack densities of 20–30 kW. Requires chilled water or refrigerant piping to the rack row.
J–M
kW (Kilowatt)
The primary unit for measuring data center power consumption. IT equipment power is measured in kW; facility power capacity is measured in kW or MW. Rack density is measured in kW per rack. PUE is calculated from kW measurements.
Latency
The time delay in data transmission, measured in milliseconds. Critical for financial trading, real-time applications, and AI inference. Colocation facilities in major markets offer low-latency connectivity to financial exchanges, cloud providers, and internet exchanges.
Load Bank
A device that simulates IT load for testing generators, UPS systems, and cooling infrastructure. Used during commissioning and annual load testing to verify that infrastructure performs as designed under actual load conditions.
Mantrap
A physical security control consisting of two interlocked doors — the second door cannot open until the first is closed and the person is authenticated. Prevents tailgating (unauthorized entry behind an authorized person). Standard in Tier III/IV data center floors.
Meet-Me Room (MMR)
A neutral space within a colocation facility where multiple carriers and customers can interconnect. The hub of the facility's connectivity ecosystem.
MTBF (Mean Time Between Failures)
The average time between equipment failures. A reliability metric — higher MTBF indicates more reliable equipment. Used in availability calculations and maintenance planning.
MTTR (Mean Time to Repair)
The average time to restore service after a failure. A maintainability metric — lower MTTR indicates faster recovery. Depends on spare parts availability, staff expertise, and documentation quality.
N–P
N+1 Redundancy
One additional component beyond the minimum required to carry the load. Any single component can fail without affecting operations. Minimum redundancy for Tier III concurrent maintainability.
PDU (Power Distribution Unit)
A device that distributes power from UPS systems to rack rows (floor-standing PDU) or from rack PDUs to individual servers (rack PDU). Intelligent rack PDUs provide per-outlet metering, remote switching, and environmental monitoring.
PUE (Power Usage Effectiveness)
Total Facility Power ÷ IT Equipment Power. The primary data center efficiency metric. A PUE of 1.0 is theoretical perfection; industry average is approximately 1.5; world-class facilities achieve 1.1–1.2.
Raised Floor
A modular floor system elevated 12–24 inches above the structural floor, providing underfloor space for power and cooling distribution. Traditional in older data centers; increasingly replaced by overhead distribution in new construction.
Rack Unit (U)
The standard unit of measurement for rack-mounted equipment height. 1U = 1.75 inches (44.45mm). A 1U server is 1.75 inches tall; a 4U server is 7 inches tall. Standard racks are 42U tall.
Q–S
RDHx (Rear-Door Heat Exchanger)
A heat exchanger that replaces the standard rear door of a server rack, capturing heat as it exits the rack. Chilled water flows through the heat exchanger, cooling exhaust air before it enters the room. Effective for 20–40 kW per rack (passive) or 40–60 kW (active).
Redundancy
Duplication of critical components or systems to increase reliability. Common redundancy levels: N+1 (one spare), 2N (full duplication), 2(N+1) (full duplication with spares). Higher redundancy increases cost and complexity but improves availability.
SLA (Service Level Agreement)
A contractual commitment by a service provider to deliver a specified level of service. Data center SLAs typically cover power availability, cooling, and network connectivity. SLA remedies are usually service credits, not actual damages.
SOC 2 Type II
A security audit standard that evaluates a service organization's controls for security, availability, processing integrity, confidentiality, and privacy over a period of time (typically 6–12 months). Required by many enterprise customers as evidence of operational security controls.
Spine-Leaf Architecture
A two-tier network architecture where leaf switches connect to servers and spine switches connect to all leaf switches. Any server can reach any other server in two hops. Provides predictable, low-latency connectivity and easy horizontal scaling.
Static Transfer Switch (STS)
A device that provides sub-cycle (less than 4ms) transfer between two power sources. Faster than mechanical ATS. Used for the most sensitive loads where even 100ms of power interruption is unacceptable.
T–Z
Thermographic Scanning
Infrared imaging of electrical equipment to identify hot spots caused by loose connections, failing components, or overloaded circuits. Should be performed annually on all critical electrical equipment as part of preventive maintenance.
Tier (Uptime Institute)
A classification system for data center infrastructure reliability. Tier I through Tier IV represent progressively higher levels of redundancy and availability. Tier III (99.982%) is the standard for most enterprise deployments; Tier IV (99.995%) is for the most critical applications.
UPS (Uninterruptible Power Supply)
A device that provides instantaneous backup power during utility outages and conditions power quality. Double-conversion (online) UPS is standard for mission-critical data centers — IT equipment always runs on UPS output, providing zero transfer time and continuous power conditioning.
VRLA (Valve-Regulated Lead-Acid)
The traditional UPS battery technology. Sealed, maintenance-free, and well-understood. Typical life: 5–7 years. Being replaced by lithium-ion in new deployments due to longer life (10–15 years) and smaller footprint.
White Space
The area of a data center floor dedicated to IT equipment (servers, storage, networking). Distinct from gray space (mechanical and electrical infrastructure) and office space. White space utilization is a key capacity metric.
Zero-Downtime Maintenance
The ability to perform maintenance on any infrastructure component without affecting IT operations. Achieved through concurrent maintainability (Tier III) — maintaining the inactive distribution path while the active path continues to supply power. Requires dual-corded IT equipment and redundant distribution paths.