Data Center Terms: Colocation and Networking Glossary

A+B power means delivering two independent power feeds (A and B) to a rack. Devices with dual power supplies connect to both feeds and keep running on one if the other is lost.

What it means for a business: For the redundancy to work, the total load must not exceed the capacity of a single feed, so each feed should be loaded to no more than half of its capacity. Single-supply devices use an automatic transfer switch (ATS).

Related: PDU · UPS · N+1 and 2N redundancy

Availability percentage is the share of a period during which a service is up and running. On a yearly basis, 99.9% allows about 8.76 hours of downtime, 99.99% about 52.6 minutes and 99.999% about 5.3 minutes.

What it means for a business: Each extra “nine” cuts allowed downtime by a factor of ten and raises cost noticeably. Set the target based on what an hour of downtime actually costs the business.

Related: downtime cost calculator · SLA · RTO

Bandwidth is the maximum amount of data a connection can carry per unit of time, expressed in Mbps or Gbps. Port speed is the physical ceiling, while committed bandwidth is the capacity that can be used continuously.

What it means for a business: Bandwidth and latency are different things; a faster link does not reduce the delay of reaching a distant server. Plan for traffic peaks where backup, replication and large file transfers take place.

Related: latency · IP transit

BGP (Border Gateway Protocol) is the routing protocol that independent networks on the internet (autonomous systems) use to announce which IP prefixes can be reached and through which path. The current version, BGP-4, is defined in RFC 4271.

What it means for a business: Organizations with their own IP block and AS number can use BGP to connect to several carriers, so traffic shifts automatically to another line if one fails. A misconfigured BGP session, on the other hand, can cause wide reachability problems.

Related: IP transit · multi-carrier connectivity

A carrier-neutral data center is not owned by a single telecom operator and hosts several carriers in the same facility. Customers can choose their connectivity provider independently of the facility owner.

What it means for a business: It makes it possible to compare carriers on price and performance, add a second line or change providers without moving any servers. That flexibility has a direct effect on long-term connectivity costs.

Related: meet-me room · multi-carrier connectivity · colocation

Colocation means placing a company’s own servers and network devices in a rented rack, or part of a rack, inside a data center. The hardware belongs to the customer; the data center provides power, cooling, physical security and connectivity infrastructure.

What it means for a business: It suits organizations that have already invested in hardware but do not want to run a server room. What you rent is space and infrastructure; hardware maintenance and operating system management stay with the customer unless agreed otherwise.

Related: what is colocation · colocation rack rental · dedicated server

A cross-connect is a physical cable (fiber or copper) linking two parties inside the same data center, for example a customer rack and a carrier’s or cloud provider’s equipment. It is usually patched through the meet-me room and billed by the data center.

What it means for a business: A cross-connect carries traffic over a private link without crossing the public internet, which keeps latency low and bandwidth predictable. Every new carrier or partner connection means another cross-connect.

Related: meet-me room · carrier-neutral data center · IP transit

A data center is a facility that houses servers, storage and network equipment with continuous power, cooling, physical security and carrier connectivity. The three main types are enterprise facilities owned by one company, multi-tenant colocation facilities and cloud provider facilities.

What it means for a business: Moving workloads to a professional data center is the most common way to take server-room tasks such as air conditioning, UPS maintenance, fire suppression and access control off the company’s plate. Location, power redundancy and connectivity options drive the choice.

Related: colocation rack rental · colocation · Tier classification

A generator is a backup power source, usually diesel, that produces electricity for the data center during long utility outages. During an outage the UPS carries the load while the generator starts within seconds to a few minutes.

What it means for a business: For long outages, what matters is the fuel stock on site, refuelling contracts and regular load testing. How often the generators are tested under load is a concrete question to ask any provider.

Related: UPS · N+1 and 2N redundancy · Tier classification

A hot/cold aisle layout arranges racks so that their fronts face each other, separating cold supply air from the hot air exhausted by equipment. Physically enclosing one of the aisles is called aisle containment.

What it means for a business: When hot and cold air mix, cooling efficiency drops and equipment runs hotter. In colocation racks, closing empty U spaces with blanking panels helps preserve the layout.

Related: PUE · rack

IP transit is a paid service in which a carrier carries a customer’s traffic to the rest of the internet and provides reachability to all internet destinations. It is usually priced on a committed bandwidth (commit) plus overage model.

What it means for a business: In colocation offers, internet access is usually sold as IP transit. Compare port speed, committed volume, 95th percentile billing and overage fees together.

Related: BGP · bandwidth · cross-connect

ISO 22301 is the international standard that sets requirements for a business continuity management system. It makes an organization’s preparation for disruptions, its response and the restart of critical activities within acceptable time frames systematic.

What it means for a business: An infrastructure provider’s ISO 22301 certification shows that it runs a planned and audited process for the continuity of its own services. It supports, but does not replace, the customer’s own business continuity plan.

Related: business continuity · certifications · disaster recovery

Latency is the time it takes a data packet to travel from source to destination, usually measured in milliseconds (ms) as round-trip time (RTT). Physical distance, the number of routers in between and network congestion determine latency.

What it means for a business: ERP, remote desktop, VoIP and synchronous replication are sensitive to latency. A data center in Türkiye, close to users and branches, improves response times for these applications.

Related: bandwidth · synchronous replication

A meet-me room (MMR) is the shared interconnection room in a data center where the physical connections of carriers, internet service providers and customers come together. Cross-connects are patched through the distribution panels in this room.

What it means for a business: The number of carriers present in a facility’s meet-me room determines how wide the connectivity options are and how easy it is to switch providers. Ask about it before choosing a rack to avoid depending on a single carrier.

Related: cross-connect · carrier-neutral data center · multi-carrier connectivity

Multi-carrier connectivity means a facility or network connects to the internet through more than one independent carrier. If one carrier fails, traffic is moved to the others, usually through BGP routing.

What it means for a business: In an infrastructure tied to a single carrier, a backbone failure at that carrier makes every service unreachable. It is also worth asking whether the connections enter the building over physically diverse paths.

Related: BGP · carrier-neutral data center · IP transit

N is the number of components a system needs to carry its load. N+1 means there is one spare component on top of that; 2N means the whole system is duplicated, so every component has an independent twin.

What it means for a business: N+1 tolerates the failure of a single component, while 2N can also survive the loss of an entire distribution path. In a proposal, check whether the claim refers to power, cooling or connectivity.

Related: Tier classification · UPS · A+B power feeds

Network segmentation divides a network into isolated parts for security and management purposes. It is implemented with VLANs, subnets, firewall rules and micro-segmentation.

What it means for a business: Segmentation makes it harder for threats such as ransomware to move laterally across the network. Separating backup and management networks from production is one of the basic measures for keeping recovery possible after an attack.

Related: VLAN · zero trust · ransomware

Out-of-band management is access to a server through a separate management interface that is independent of the operating system and the main network connection. IPMI is an open standard; iDRAC is the Dell equivalent and iLO the HPE equivalent.

What it means for a business: It provides remote console access, power control and hardware health data even when the server will not boot or the operating system has crashed. These interfaces should never be exposed to the internet and should be reachable only through a VPN or a separate management network.

Related: remote hands · VPN · dedicated server rental

A PDU (power distribution unit) is a multi-outlet unit that distributes incoming power to the devices in a rack. Basic models only distribute power; metered and switched models show consumption per outlet and allow outlets to be turned on and off remotely.

What it means for a business: A metered PDU shows how close the rack is to its power limit. When dual-supply servers are connected to two separate PDUs, the failure of one PDU does not take the server down.

Related: A+B power feeds · power per rack · out-of-band management

Power per rack is the electrical capacity, in kilowatts (kW), allocated to a rack that the equipment can draw continuously. The number of devices that physically fit in a rack is often different from the number the allocated power can support.

What it means for a business: GPU servers and dense storage systems need a lot of power; once the power limit is reached, no more devices can be added even if the rack is half empty. Use measured consumption rather than nameplate ratings for a realistic figure.

Related: rack power calculator · PDU · PUE

PUE is an efficiency ratio calculated by dividing a data center’s total facility energy by the energy used by IT equipment. The closer the value is to 1.0, the less energy goes to non-IT loads such as cooling, lighting and power conversion. The metric was defined by The Green Grid and later standardized as ISO/IEC 30134-2.

What it means for a business: PUE describes energy efficiency; on its own it says nothing about reliability or outage risk. It is tracked where energy costs are passed through in pricing and in sustainability reporting.

Related: power per rack · hot/cold aisle

A rack is a metal cabinet in which servers and network devices are mounted horizontally on standard 19-inch rails. Its height is measured in rack units (U); 42U and 47U racks are common in data centers.

What it means for a business: Rack offers come as a full rack, a half rack or a set number of U. When comparing them, read the height together with the power allocated to the rack, the number of power feeds and the locked-access arrangements.

Related: rack unit (U) · power per rack · colocation rack rental

A rack unit (U) is the unit of vertical space inside a rack: 1U equals 1.75 inches, or 44.45 mm. Servers are built in heights such as 1U, 2U or 4U, so a 42U rack can in theory hold 42 devices of 1U each.

What it means for a business: In practice the whole rack is never available for servers: patch panels, switches, cable management and PDUs take space too. When requesting a quote, add up the U of current equipment and leave room for growth.

Related: rack · rack power calculator

Remote hands is a service in which data center staff physically work on the customer’s equipment in the rack on the customer’s behalf. Typical tasks are rebooting a server, replacing a cable, inserting a disk and checking the console screen.

What it means for a business: When servers are in another city, there is no need to send a team for a small intervention. The scope, response time and billing terms of the service should be written clearly into the contract.

Related: out-of-band management · colocation · managed services

A site-to-site VPN joins the networks of two locations, such as a head office and servers in a data center, through an always-on encrypted tunnel. The tunnel is usually built with IPsec between the firewalls or routers at each end.

What it means for a business: Users reach servers in the data center over private IP addresses from the office without running any extra client. Tunnel performance depends on the internet connections at both ends.

Related: VPN · firewall solutions

An SLA (service level agreement) is a contract that sets a provider’s measurable targets, such as availability, response time and support scope. It also states the measurement period, how planned maintenance is counted and what applies when a target is missed.

What it means for a business: When comparing two SLAs, read the percentage together with the measurement period (monthly or yearly), the exclusions and the first-response times. A high percentage with a narrow scope means little on its own.

Related: MAV Cloud SLA · availability percentage

The Tier classification is the Uptime Institute system that rates data center infrastructure on four levels. Tier I is basic capacity, Tier II adds redundant capacity components, Tier III is concurrently maintainable and Tier IV is fault tolerant.

What it means for a business: “Designed to Tier III” is not the same as a Tier certification awarded by Uptime Institute. A certification claim can be checked against the certified facility records Uptime Institute publishes.

Related: N+1 and 2N redundancy · availability percentage · data center

A UPS (uninterruptible power supply) provides power from batteries when mains power fails or fluctuates. In a data center, the UPS bridges the gap until the generators start and stabilize.

What it means for a business: Office-grade UPS units last only minutes and can fail at the first outage if battery maintenance is neglected. Redundant UPS infrastructure in a data center removes that maintenance burden from the business.

Related: generator · A+B power feeds · N+1 and 2N redundancy

A VLAN (virtual LAN) creates logically separated networks on the same physical switch infrastructure. The VLAN tag added to Ethernet frames is defined in the IEEE 802.1Q standard.

What it means for a business: Keeping server, management, backup and guest traffic in separate VLANs limits how far a problem in one segment can spread. Traffic between VLANs should be controlled by firewall rules.

Related: network segmentation · firewall

A VPN (virtual private network) encrypts traffic between two points to create a private tunnel over a shared network such as the internet. A remote-access VPN connects a single user, while a site-to-site VPN connects two networks.

What it means for a business: VPN is a common way to give remote staff secure access to company resources. It is not enough on its own: VPN logins should be protected with multi-factor authentication and access limited by role.

Related: site-to-site VPN · MFA · zero trust network security