Data Centre Glossary: Key Terms Explained

Data centre and connectivity terminology can be confusing, especially for organisations evaluating a provider for the first time. This glossary explains the terms that matter most when choosing a colocation or interconnection partner in Africa, in plain language.

A data centre built with the power density, cooling infrastructure and low-latency connectivity required to support AI, machine learning and high-performance computing workloads, beyond what standard colocation facilities were originally designed for.

The routing protocol networks use to exchange traffic directly with each other, commonly used for peering within a data centre without needing extra physical cabling for each connection.

Renting virtual computing resources from a provider, with no physical hardware of your own to manage, as opposed to colocation, where you own and manage your own physical servers.

Renting virtual computing resources from a provider, with no physical hardware of your own to manage, as opposed to colocation, where you own and manage your own physical servers.

Placing your own physical servers and hardware within a secure, third-party facility, which provides the power, cooling, connectivity and physical security around your equipment.

A direct physical cable connection between two clients within the same data centre, allowing them to exchange data privately and directly without routing traffic externally.

Keeping data storage and processing within a specific country or region's legal jurisdiction, an increasingly important requirement for regulated industries and government applications.

The combined network of carriers, cloud providers, internet exchanges, enterprises and platforms that interconnect within a data centre, creating value beyond the physical infrastructure itself.

A smaller facility positioned closer to end users than a core data centre, reducing latency for applications such as digital banking, content delivery and real-time analytics.

A licensed operator that builds and manages fibre-optic infrastructure, providing wholesale or retail connectivity services over that network.

A fibre-optic connection that runs directly to an individual residence or business premises, offering higher bandwidth than legacy copper connections.

Data centre infrastructure built at a scale designed to support very large computing demands, typically used by major cloud providers and large enterprises with significant capacity requirements.

The physical and virtual connections that allow different networks, cloud platforms and enterprises to exchange data directly within a facility, improving performance and reducing cost compared with routing traffic externally.

A physical location where multiple networks connect to exchange internet traffic directly with each other, rather than routing through third-party networks.

An internationally recognised certification for information security management systems, verifying that a facility follows structured, audited security practices.

The time it takes for data to travel between two points on a network. Lower latency generally means faster, more responsive applications, which is why proximity to subsea cables, internet exchanges and end users matters.

Localised, scheduled power cuts to specific overloaded suburbs or feeders in South Africa, distinct from nationwide load shedding, and which can occur even when load shedding is not in effect.

A nationwide, centrally coordinated measure used in South Africa to reduce demand on the electricity grid when supply cannot meet demand, implemented in stages and announced in advance.

A dedicated space within a data centre where multiple carriers and networks physically connect to each other, forming the technical hub of a carrier-neutral facility's interconnection capability.

A power or cooling design that includes one additional unit of capacity beyond what is minimally required, so that a single failure does not interrupt operations.

An operating model where a data centre brings together carriers, cloud providers, internet exchanges and enterprises within one environment, with no party given preferential or exclusive control over how others connect. Broader than carrier neutrality, which applies specifically to carriers.

An arrangement between two networks to exchange traffic directly with each other, typically to reduce transit costs and improve performance.

A physical location where a network or service provider has equipment to connect to other networks, often used to extend reach into a new city or region.

A standard unit of measurement for server rack space, with one rack unit equal to 1.75 inches of vertical height. Colocation is often sold in quarter-rack, half-rack or full-rack increments.

On-site technical support provided by a data centre's own team, allowing clients to manage routine tasks on their equipment without travelling to the facility themselves.

An undersea fibre-optic cable that carries data between continents, forming the backbone of international internet connectivity. Facilities located at or near a subsea cable landing station typically offer lower latency for international traffic.

A data centre design standard, certified by the Uptime Institute, indicating a facility is built with redundant infrastructure components allowing maintenance to be performed without disrupting operations.

The percentage of time a facility's infrastructure remains operational and available, commonly cited as a headline reliability metric (for example, 99.98% uptime).

The usable technical floor space within a data centre available for client racks and equipment, distinct from the facility's total building footprint.