Colocation Glossary - Data Center Terms & Definitions | Arizona Colocation

Colocation & Data Center Glossary

Definitions for 68+ industry terms to help you understand colocation and data center technology.

2 A B C D E F G H I K L M N P R S T U V W

2

2N Redundancy: A redundancy model where the entire system is fully duplicated. If a system needs 3 units, 2N means 6 are installed in two completely independent paths. This is the standard for Tier 4 data centers.

A

A+B Power: A dual-feed power configuration providing two independent electrical paths to each rack. If one feed fails, the other continues supplying power. A+B power is standard in Tier 3+ data centers and essential for mission-critical deployments.
ATS (Automatic Transfer Switch): A device that automatically switches power from the primary utility feed to a backup generator when it detects a power failure. ATS ensures seamless failover with minimal interruption, typically within seconds.

B

Bandwidth: The maximum rate of data transfer across a network, measured in bits per second (bps). In colocation, bandwidth refers to the internet connectivity provided to your equipment. Common options include 100 Mbps, 1 Gbps, and 10 Gbps.
Bare Metal Server: A physical server dedicated entirely to a single tenant, as opposed to a virtual machine sharing hardware with others. Colocation is inherently bare metal since you own the hardware. Bare metal offers maximum performance with no hypervisor overhead.
BGP (Border Gateway Protocol): The routing protocol that manages how data travels across the internet between autonomous systems. In colocation, BGP allows you to advertise your own IP addresses and control traffic routing across multiple carriers for redundancy and performance.
Blended IP Transit: Internet connectivity that routes traffic across multiple upstream providers for optimal performance and redundancy. Blended transit is the most common bandwidth option in colocation and provides a balance of cost and quality.
BSA (Building Security Act / Building Standards): Building codes and standards that govern data center construction, including fire suppression, structural integrity, and electrical systems. Data centers must comply with local BSA requirements, NFPA codes, and building standards to ensure facility safety and insurance compliance.
Burstable Bandwidth: A bandwidth model where you pay for a base commitment but can temporarily exceed it when needed. Billing is typically based on the 95th percentile of usage, meaning the top 5% of traffic spikes are excluded from billing calculations.

C

Cabinet: An enclosed rack unit used to house servers and networking equipment. Standard cabinets are 42U tall, 19 inches wide, and provide physical security with locking doors.
Cage: A fenced enclosure within a data center that provides a private, lockable area containing multiple racks. Cages offer enhanced security and privacy compared to shared rack space.
Carrier Hotel: A building that serves as a major interconnection point for telecommunications carriers. Carrier hotels contain meet-me rooms where multiple network providers can connect and exchange traffic.
Carrier Neutral: A data center that allows tenants to choose from multiple network carriers and ISPs, rather than being locked into a single provider. Carrier neutrality ensures competitive pricing and diverse connectivity options.
CDN (Content Delivery Network): A geographically distributed network of servers that caches and delivers content to users from the nearest location. Many colocation providers offer direct connections to major CDNs like Cloudflare, Akamai, and Fastly for improved content delivery.
Cloud On-Ramp: A direct, private connection from a colocation facility to a public cloud provider (AWS, Azure, GCP). Cloud on-ramps bypass the public internet for lower latency, higher bandwidth, and reduced data transfer costs in hybrid deployments.
Colocation (Colo): A data center service where businesses rent space to house their own servers and networking equipment. The provider supplies physical space, power, cooling, bandwidth, and security.
Compliance: Adherence to regulatory requirements and industry standards governing data handling, security, and privacy. Common colocation compliance standards include SOC 2, HIPAA, PCI DSS, ISO 27001, and NIST 800-53.
Concurrently Maintainable: A facility design where any component can be maintained or replaced without taking the system offline. This is a key characteristic of Tier 3 and above data centers.
Cross-Connect: A physical cable connecting your equipment to another tenant, carrier, or cloud provider within the same data center. Cross-connects provide the lowest possible latency and typically cost $100-300/month each.
CRAC Unit: Computer Room Air Conditioning. Precision cooling units specifically designed for data center environments. CRAC units maintain tight temperature and humidity tolerances to protect equipment.

D

DDoS Protection: Distributed Denial of Service mitigation. Security services that detect and filter malicious traffic designed to overwhelm your network. Many colocation providers offer DDoS protection as an add-on service.
Dedicated Server: A physical server leased entirely to one customer. Unlike colocation where you own the hardware, a dedicated server is owned and maintained by the hosting provider. You get root access but don't physically manage the machine.
Disaster Recovery (DR): A strategy and set of procedures for restoring IT infrastructure and data after a catastrophic event. Arizona is a popular DR location due to its low natural disaster risk, geographic distance from coastal threats, and affordable colocation pricing.
Diverse Path: Network or power connections that follow physically separate routes to eliminate single points of failure. Diverse paths ensure that cutting one cable or losing one utility feed does not cause a complete outage.

E

Edge Computing: Processing data closer to where it is generated rather than in a centralized data center. Edge colocation facilities are smaller, distributed locations that reduce latency for applications like IoT, gaming, and autonomous vehicles.
Egress: Outbound data transfer from a data center or cloud environment. Egress fees can be a significant cost in cloud hosting but are typically included or much cheaper in colocation bandwidth plans.

F

Fault Tolerant: A system design that can continue operating without interruption when a component fails. Tier 4 data centers are fully fault tolerant with 2N redundancy across all systems.
Fiber Optic: High-speed data transmission cables that use light pulses through glass or plastic strands. Fiber provides much higher bandwidth and lower latency than copper cabling. Single-mode fiber supports distances over 100km, while multi-mode is used for shorter runs within facilities.
Free Cooling (Economizer): A cooling method that uses outside air or evaporative cooling instead of mechanical refrigeration when ambient conditions permit. Arizona's dry climate enables free cooling for much of the year, reducing PUE.

G

GPU Colocation: Colocation services specifically designed for high-density GPU servers used in AI/ML training, rendering, and high-performance computing. GPU racks require significantly more power (often 20-40+ kW per rack) and cooling than standard servers.

H

HIPAA: Health Insurance Portability and Accountability Act. A US federal law requiring safeguards for protected health information (PHI). HIPAA-compliant colocation facilities must provide physical security, access controls, audit logging, and sign a Business Associate Agreement (BAA).
Hot Aisle / Cold Aisle: A rack layout strategy where racks face each other in alternating rows. Equipment intakes face the cold aisle and exhausts face the hot aisle, improving cooling efficiency by preventing hot and cold air from mixing.
HVAC: Heating, Ventilation, and Air Conditioning. The climate control systems responsible for maintaining optimal temperature and humidity in a data center, typically 68-72°F and 40-60% relative humidity.
Hybrid Cloud: An IT architecture combining on-premises or colocation infrastructure with public cloud services. Hybrid cloud allows businesses to keep sensitive or steady workloads in colocation while using cloud for burst capacity and SaaS applications.
Hyperscale: Extremely large data center deployments, typically 100,000+ sq ft with 10+ MW of power capacity. Hyperscale facilities are built for massive cloud providers, AI companies, and enterprises needing enormous compute and storage capacity.
Hypervisor: Software that creates and manages virtual machines on physical hardware. Type 1 hypervisors (VMware ESXi, KVM) run directly on hardware. Type 2 hypervisors (VirtualBox) run on top of an operating system. Used in VPS hosting but not in traditional colocation.

I

IPMI / iLO / iDRAC: Out-of-band management interfaces that allow remote access to server hardware independent of the operating system. IPMI is the standard protocol; HP calls theirs iLO, Dell calls theirs iDRAC. Essential for managing colocation servers remotely.
ISO 27001: An international standard for information security management systems (ISMS). ISO 27001 certification demonstrates that a data center has implemented comprehensive security controls covering people, processes, and technology.
IX (Internet Exchange): A physical location where multiple internet service providers and networks exchange traffic via peering. Being in or near an IX reduces latency and transit costs. Phoenix has several active internet exchanges.

K

kW (Kilowatt): A unit of power measurement commonly used in colocation pricing. Power allocation per rack typically ranges from 2 kW (low density) to 30+ kW (high density). Power is often the largest variable cost in colocation.

L

Latency: The time delay between sending a request and receiving a response, measured in milliseconds (ms). Lower latency means faster communication. Arizona provides <8ms latency to Los Angeles and <15ms to Dallas.
Liquid Cooling: A cooling method that uses liquid (water or specialized coolant) to remove heat directly from server components. More efficient than air cooling for high-density AI/GPU workloads. Methods include direct-to-chip, immersion cooling, and rear-door heat exchangers.

M

Man Trap: A two-door security vestibule where the first door must close and lock before the second door will open. Man traps prevent unauthorized individuals from following an authorized person into a secure area.
Managed Hosting: A service where the provider owns and manages the server hardware, operating system, and sometimes applications on your behalf. Managed hosting sits between colocation (you manage everything) and cloud (provider manages everything).
Meet-Me Room: A dedicated space within a data center where network carriers terminate their connections. The meet-me room enables cross-connects between tenants and carriers.
MW (Megawatt): A unit of power equal to 1,000 kilowatts. Used to describe the total power capacity of large data center facilities. Arizona's major data centers range from 5 MW to 240+ MW of total campus capacity.

N

N+1 Redundancy: A redundancy model where one additional unit exists beyond the minimum required. If a system needs 3 cooling units to operate, N+1 means 4 are installed. If one fails, the remaining 3 handle the full load.
NIST 800-53: A comprehensive security and privacy controls framework published by the National Institute of Standards and Technology. Required for federal government systems and widely adopted by enterprise data centers as a security baseline.

P

PCI DSS: Payment Card Industry Data Security Standard. A set of security requirements for organizations that handle credit card data. PCI DSS-compliant colocation provides the physical security controls needed for cardholder data environments.
PDU (Power Distribution Unit): A device that distributes electrical power to servers and equipment within a rack. Intelligent PDUs offer remote monitoring and per-outlet switching. Redundant PDUs (A+B) connect to separate power feeds for failover protection.
Peering: A network arrangement where two internet service providers exchange traffic directly, bypassing the need to route through a third party. Peering reduces latency and costs.
PUE (Power Usage Effectiveness): A metric measuring data center energy efficiency. PUE = Total Facility Power / IT Equipment Power. A PUE of 1.0 is perfect efficiency. Industry average is ~1.6. Arizona facilities commonly achieve 1.2-1.4 due to the dry climate.

R

Rack Unit (U): The standard unit of measurement for server rack space. One U equals 1.75 inches (44.45mm) of vertical space. Standard racks are 42U tall. A typical 1U server is the thinnest standard form factor.
RAID: Redundant Array of Independent Disks. A storage technology that combines multiple hard drives for performance and/or redundancy. Common levels include RAID 1 (mirroring), RAID 5 (striping with parity), and RAID 10 (mirrored stripes). Critical for data protection in colocation servers.
Remote Hands: A service where data center technicians perform physical tasks on your behalf, such as rebooting servers, swapping drives, or adjusting cabling. Remote hands are typically available 24/7 and billed per incident or included in your plan.
RTO / RPO: Recovery Time Objective (RTO) is the maximum acceptable downtime after a failure. Recovery Point Objective (RPO) is the maximum acceptable data loss measured in time. Together they define disaster recovery requirements and drive colocation DR planning.

S

SLA (Service Level Agreement): A contractual commitment defining guaranteed uptime, support response times, and compensation for service failures. Standard colocation SLAs guarantee 99.99% or higher uptime.
SOC 2 Type II: A compliance certification that verifies a service provider has implemented effective security controls over an extended period. SOC 2 Type II covers security, availability, processing integrity, confidentiality, and privacy.
Structured Cabling: A standardized approach to designing and installing cable infrastructure in a data center. Structured cabling uses organized pathways, patch panels, and labeling systems to ensure maintainability, scalability, and airflow management.
Suite: A private, fully enclosed room within a data center, larger than a cage. Suites offer the highest level of privacy and security for enterprise customers and can include custom power, cooling, and access controls.

T

Tier Classification: The Uptime Institute's four-level rating system for data center infrastructure. Tier 1 (basic, 99.671% uptime) through Tier 4 (fault tolerant, 99.995% uptime). Each tier adds redundancy and concurrent maintainability.
TOR Switch: Top-of-Rack switch. A network switch mounted at the top of a server rack that connects all servers in that rack to the broader network. TOR architecture simplifies cabling and is the most common data center network design.

U

UPS (Uninterruptible Power Supply): Battery backup systems that provide continuous power during the transition from utility power to generator power during an outage. UPS systems also condition power to protect equipment from surges and sags.
Uptime SLA: The guaranteed percentage of time a data center will be operational, typically expressed as a percentage like 99.99% (52.6 minutes of downtime per year) or 99.999% (5.26 minutes per year). SLA credits compensate customers if uptime falls below the guarantee.

V

Virtualization: Technology that creates virtual versions of physical hardware, allowing multiple operating systems and applications to run on a single physical server. Common platforms include VMware, KVM, and Hyper-V. VPS hosting relies on virtualization, while colocation uses physical (bare-metal) servers.
VLAN: Virtual Local Area Network. A logical network segmentation that isolates traffic within a shared physical network. VLANs are used in colocation to separate customer traffic, improve security, and optimize network performance.
VPS (Virtual Private Server): A virtual machine sold as a service by a hosting provider. Unlike colocation where you own hardware, a VPS runs on shared physical servers using virtualization technology. VPS offers lower upfront costs and instant provisioning, while colocation provides dedicated performance and full hardware control.

W

Wholesale Colocation: Large-scale colocation where a customer leases entire data halls, floors, or buildings rather than individual racks. Wholesale deals typically involve 500 kW+ of power and multi-year contracts at lower per-kW pricing than retail colocation.