Edge Computing Glossary: Key Terms and Definitions

3G, 4G, 5G

Third‑, fourth‑, and fifth‑generation cellular technologies. 3G introduced smartphones and mobile browsers; 4G delivers true broadband access; 5G offers massive bandwidth and low latency, enabling use cases from smartphones to autonomous vehicles, with edge computing regarded as a key 5G component.

Access Edge

The sub‑layer of the service‑provider edge closest to the physical last‑mile network, often located at a radio access node (RAN) or cable head‑end. It consists of highly distributed server‑grade infrastructure (e.g., at cell towers, cable distribution points, central offices) and typically serves workloads requiring ultra‑low latency (1 ms–30 ms) within a 15 km radius of the radio or cable head‑end.

Access Network

The network that connects users and devices to their local service‑provider network, contrasting with the core network that interconnects service providers.

Aggregation Edge

The service‑provider layer one hop away from the access edge, which can be a single medium‑sized data center or a collection of micro‑data centers forming a hierarchical topology between regional and access edges.

Base Station

An RAN element that transmits and receives radio signals for user equipment within one or more cells. It may use integrated antennas or feed‑line cables to antenna arrays and can be split into software‑based functional blocks for flexibility, cost, and performance.

Baseband Unit (BBU)

A base‑station component responsible for baseband radio signal processing, typically implemented in dedicated hardware. In C‑RAN architectures, BBU functions can run as virtual network functions (VNFs) in software.

Central Office (CO)

A geographic aggregation point for telecom infrastructure. Physically designed for equipment housing, it usually lacks the space, power, cooling, and fire‑suppression capacity required for full‑scale edge data‑center workloads unless specially provisioned.

CO Redesigned as Data Center (CORD)

Deploys data‑center‑grade compute and storage within a central office.

Centralized Data Center

A large, often hyperscale, physical and logical entity that houses massive compute, storage, and networking resources, typically serving many tenants and located far from end users, forming the backbone of cloud computing.

Cloud Computing

A system that provides on‑demand access to shared pools of compute, storage, and networking resources, usually realized through a few large centralized data centers and regional data centers.

Cloud‑Native Network Function (CNF)

A cloud‑native application that implements network functions. A CNF consists of one or more microservices built with cloud‑native principles such as immutable infrastructure, declarative APIs, and repeatable deployment processes. Examples include packet filters and firewalls composed of multiple microservices.

Cloud Node

A compute node—single server or a group of resources—operating as part of cloud infrastructure, typically residing inside a centralized data center.

Cloud RAN (C‑RAN)

An evolution of RAN that separates wireless base‑station functions into a remote radio head (RRH) and a centralized BBU. BBUs can be placed at aggregation points (e.g., distributed antenna system hubs) to improve efficiency and enable migration toward cloud‑based RAN.

Cloud Service Provider (CSP)

A large organization that operates centralized and regional data centers, commonly delivering public‑cloud services. Also referred to as a cloud service operator (CSO).

Cloudlet

In academia, a cloudlet is an edge‑enhanced public or private cloud located at the infrastructure edge, synonymous with edge cloud. It represents the middle tier (Tier 2) in a three‑tier architecture, with Tier 1 being the cloud and Tier 3 comprising smartphones, wearables, and sensors.

Colocation

The practice of deploying computing, storage, and networking infrastructure owned by different parties in the same physical location. Unlike shared infrastructure, colocation does not require multiple tenants within the edge data center.

Computational Offloading

An edge‑computing use case where tasks are offloaded from edge devices to infrastructure edge for remote processing, aiming to reduce execution latency and device energy consumption. Also known as cloud offload or cyber foraging.

Content Delivery Network (CDN)

A distributed system that places content (e.g., streaming video) closer to users, reducing network congestion and improving response time. When combined with edge computing, CDN nodes run within edge‑data‑center software.

Core Network

The service‑provider network layer that interconnects access networks and devices with other operators and services, enabling data exchange across the Internet. It may be several hops away from edge resources.

CPE (Customer‑Premises Equipment)

A device that receives mobile signals and forwards them as Wi‑Fi, such as a wired modem, allowing users to connect to the service‑provider access network—typically one hop from the infrastructure edge.

Data Center

A facility that houses multiple high‑performance compute and storage nodes, consolidating large amounts of resources in one location. Scales range from centralized to regional to edge data centers.

Data Gravity

The phenomenon where data becomes increasingly difficult to move as its volume grows and the distance to processing endpoints increases, causing applications to gravitate toward the data’s location.

Data Ingest

The process of receiving large volumes of data for storage and later processing, such as ingesting video streams in an edge data center for subsequent analysis.

Data Reduction

Using an intermediate point between data producers and consumers to intelligently reduce the amount of data transmitted without losing its meaning, e.g., smart deduplication.

Data Sovereignty

The concept that data is subject to the laws and regulations of the country, state, or industry where it resides.

Device Edge

The edge‑computing capability located at the last‑mile device or user side, typically relying on gateways or local devices to collect and process data. It differs from infrastructure edge, which uses provider‑owned resources.

Device Edge Cloud

An extension of edge‑cloud where certain workloads run on resources available at the device edge, offering low‑latency execution without full cloud elasticity.

Distributed Antenna System (DAS) Hub

A location that aggregates many radio devices, often used to support cellular networks and may connect to edge data centers.

Edge Cloud

Cloud‑like functionality deployed at the infrastructure edge, providing elastic compute, storage, and networking resources to users as an extension of centralized public or private clouds.

Edge Computing

Computing performed near the source of data or devices, integrating networking, compute, storage, and application capabilities into an open platform to reduce latency and bandwidth constraints.

Edge Data Center

A smaller, highly distributed data center placed as close as possible to the network edge. Though smaller than centralized facilities, it can perform the same functions and often operates autonomously, supporting workloads with latency requirements of 1 ms–30 ms.

Edge Exchange

A pre‑Internet traffic exchange that occurs within an edge data center, typically in the edge meet‑me‑room, and can interconnect with traditional Internet exchange points to improve end‑to‑end latency.

Edge Meet‑Me Room

A space inside an edge data center where tenants and telecom providers interconnect with each other and with other edge data centers, similar to a traditional meet‑me‑room.

Edge Network Fabric

A network interconnect system—often dark or lit fiber—that links edge data centers with other local infrastructure, effectively acting as a metropolitan‑area network.

Edge Node

A compute node (single server or a set of resources) that operates as part of edge infrastructure, typically located within an edge data center and physically closer to end users than centralized cloud nodes.

Edge‑Enhanced Application

An application that can run in a centralized data center but gains performance or functional benefits (usually latency) when deployed on edge resources.

Edge‑Native Application

An application built from the ground up to exploit edge capabilities, using cloud‑native principles while accounting for edge constraints such as limited resources, security, latency, and autonomy.

Fog Computing

An early term for edge computing that positioned compute, storage, and applications between users and the cloud to improve performance and redundancy. The term, coined by Cisco, is now largely deprecated.

Gateway Device

A device at the user edge that aggregates and forwards data from local devices (often battery‑powered) to a local data center or across the last‑mile network.

Hard Real Time

Workloads that require deterministic responses; missed deadlines can cause severe or life‑threatening failures (e.g., PLCs, vehicle brakes). Typically executed at the user edge because they cannot rely on network latency.

Infrastructure Edge

Now referred to as Service‑Provider Edge in the LF Edge taxonomy; it denotes compute resources deployed as one or more edge data centers on the provider side of the last‑mile network, offering cloud‑like capabilities with lower latency and cost.

Interconnect

Physical or logical connections—usually fiber—linking networks, data centers, or tenants, such as those found in an edge meet‑me‑room.

Internet Edge

A sub‑layer of the infrastructure edge that sits between the edge and the broader Internet, containing edge meet‑me‑rooms and other equipment to provide high‑performance interconnect.

Internet Exchange Point (IXP)

A centralized platform where different telecom operators interconnect their networks to exchange traffic, often linked to the internet edge.

IoT Edge

A subset of the smart‑device edge focused on headless compute resources for IoT use cases.

IP Aggregation

Early processing of data received from the RAN at the infrastructure edge to improve QoS before routing to core networks.

Jitter

The variation in network latency observed over time, measured in milliseconds, and critical for real‑time applications such as VoIP, autonomous driving, and online gaming.

Last Mile

The portion of the telecom network that connects service providers to end customers, influencing performance and service quality.

Latency

The time taken for a data unit to travel from source to destination, measured in milliseconds; a key metric for user experience.

Latency‑Critical Application

Applications that fail or crash if latency exceeds a threshold, such as autonomous‑vehicle control systems.

Latency‑Sensitive Application

Applications that can tolerate higher latency without failure but may experience degraded user experience (e.g., image processing).

Local Breakout

The ability to route internet‑bound traffic directly from an edge node to the Internet without traversing longer paths to centralized facilities.

Location Awareness

Using RAN data and other sources to precisely determine a user’s current and near‑future location for optimal workload placement.

Location‑Based Node Selection

Selecting the optimal edge node for a workload based on the physical proximity of the node to the device, improving performance.

MANO (Management and Orchestration)

In edge computing, the set of processes that manage and orchestrate edge devices and applications throughout their lifecycle, including configuration, monitoring, updates, and security.

Micro‑Modular Data Center (MMDC)

A small‑scale, modular data center (typically 50–150 kW) that can be deployed indoors or outdoors and combined with other data centers to increase regional capacity.

Mixed‑Criticality Workload Integration

The practice of consolidating hard‑real‑time or latency‑critical workloads with soft‑real‑time or latency‑sensitive workloads (e.g., AI/ML models) on the same edge infrastructure.

Mobile Edge

A combination of infrastructure edge, device edge, and network slicing tailored for use cases such as autonomous‑vehicle control, navigation, and in‑vehicle entertainment.

Mobile Network Operator (MNO)

The entity that owns and operates the cellular network infrastructure, including RAN equipment and edge data centers.

Mobile Virtual Network Operator (MVNO)

A service provider that offers mobile services without owning the underlying cellular infrastructure; may lease space in edge data centers.

Modular Data Center (MDC)

A portable data‑center deployment method where compute, storage, and networking are housed in transportable structures.

Multi‑Access Edge Computing (MEC)

An ETSI‑sponsored open framework that tightly couples services with the RAN, enabling applications such as video optimization, caching, and real‑time transcoding at the edge.

Near Real Time

Workloads that benefit from low latency but can tolerate some delay, unlike hard real‑time requirements.

Network Functions Virtualization (NFV)

The concept of implementing network functions as software‑based VNFs on standard x86/ARM servers, often running in edge data centers.

Network Hop

A point where data is routed or switched in a network; reducing hops is a primary performance goal of edge computing.

North‑South / East‑West Data Flow

North‑south denotes data moving upstream (e.g., from edge to cloud), south‑north denotes downstream flow, while east‑west describes lateral communication between peer resources within the continuum.

On‑Premises Data Center Edge

A sub‑category of user edge consisting of server‑grade compute located within or near a customer’s building (e.g., office or factory).

Over‑the‑Top (OTT) Service Provider

An application or service provider that does not own the underlying network but delivers content (e.g., video streaming) over it.

Perishable Data

Data that is most valuable when acted upon immediately; after processing, it may be discarded to reduce transmission costs.

Point of Presence (PoP)

A physical location where a service provider allows users or partners to connect to its network, often co‑located with edge meet‑me‑rooms.

QoE (Quality of Experience)

An advanced use of QoS metrics that measures user‑perceived performance and can dynamically adjust configurations to improve experience.

QoS (Quality of Service)

Metrics such as throughput, latency, and jitter that quantify the performance of network and data‑center services for specific applications.

Radio Access Network (RAN)

The wireless variant of the access network, typically referring to cellular networks (3G/4G/5G). 5G RAN leverages NFV and C‑RAN to support edge computing.

Real Time

Workloads that benefit from or require discrete, low‑latency timing.

Regional Data Center

A data center sized between centralized and micro‑modular facilities, positioned to serve an entire region with latency typically between 30 ms and 100 ms.

Regional Edge

A sub‑category of service‑provider edge consisting of server‑grade infrastructure in regional data centers, supporting workloads with 30 ms–100 ms latency.

Resource‑Constrained Device

A device at the device edge, often battery‑powered, with limited compute and storage, typically connected to a gateway for data forwarding.

Service Provider

An organization that offers network access to customers, usually via the last‑mile network.

Service‑Provider Edge

One of the two primary layers in the LF Edge taxonomy, representing compute deployed on the provider side of the last‑mile network, subdivided into access and regional edges.

Shared Infrastructure

Multiple parties using the same compute, storage, and networking resources within a single edge data center.

Smart Device Edge

A user‑edge sub‑category consisting of compute hardware outside a secure data center but capable of running virtualized or containerized workloads (e.g., smartphones, PCs, IoT devices).

Soft PLC

A virtualized programmable logic controller that can be integrated with other virtualized or containerized applications on general‑purpose edge infrastructure.

Soft Real Time

Workloads that benefit from low latency (e.g., video streaming) but can tolerate occasional delays without catastrophic failure.

Throughput

The amount of data transferred per second between endpoints, typically measured in bits per second.

Thick Compute

High‑end gateway or server‑grade compute located at the device edge or on‑premises edge, often housed within secure data centers.

Thin Compute

Limited‑capacity edge compute (e.g., gateways, routers) that provides minimal processing and is usually paired with thick compute resources.

Tiny ML

Deploying lightweight machine‑learning inference models on microcontroller‑based devices at the device edge (e.g., wake‑word detection on a smart speaker).

Traffic Offload

Redirecting inefficiently transmitted data to a more local destination (e.g., CDN cache) to reduce cost and improve performance; local breakout is an example.

Truck Roll

Dispatching personnel to an edge location (such as an edge data center) to resolve issues, often costly due to remote or inaccessible sites.

User Edge

The edge‑computing capability deployed on the user side of the last‑mile network, encompassing device edge, on‑premises edge, and other local resources.

V2X

An umbrella term for vehicle‑to‑vehicle, vehicle‑to‑infrastructure, and infrastructure‑to‑infrastructure communications, enabling real‑time traffic information and safety features.

V2I

Vehicle‑to‑infrastructure communication, often leveraging 5G or Wi‑Fi 6, to connect vehicles with edge data‑center‑hosted vision and path‑finding applications.

Virtual Data Center

A logical entity composed of multiple physical edge data centers, presented as a single resource pool for workload placement, redundancy, and elasticity.

Virtual Network Function (VNF)

A software‑based network function that runs on generic compute resources, often deployed in edge data centers.

Workload Orchestration

An intelligent system that determines the optimal location, timing, and priority for executing application workloads across centralized, regional, infrastructure, and device edges.

Haul

High‑speed interconnects linking two or more networks or data‑center infrastructures.