Skip to main content
Definition Website Computer and Technology

Edge Computing

Subject: Computer and Technology noun /ˈɛdʒ kəmˌpjuː.tɪŋ/ 1 definition

  1. Definition (noun)

    Entry #129 · Updated

    Edge Computing is a distributed computing paradigm that brings computation, data filtering, and storage physically closer to the devices, sensors, and users generating the data. By executing workloads on local gateways, on-premises servers, or regional network nodes instead of transmitting every raw payload across wide-area networks to a centralized cloud data center, it reduces round-trip latency, conserves upstream bandwidth, and sustains critical operations during intermittent connectivity. Rather than replacing cloud infrastructure outright, it establishes an execution boundary at the network periphery for real-time responsiveness, local data sovereignty, and pre-aggregation before selective synchronization upstream.

    Common Forms

    • Device (Far) Edge: Executes lightweight inference, signal filtering, or control loops directly on endpoint hardware—such as industrial controllers, autonomous vehicles, and smart cameras—to achieve minimal latency without network reliance.

    • On-Premises (Local) Edge: Runs dedicated gateways, ruggedized servers, or micro data centers inside facilities like factories, hospitals, and retail stores to aggregate local device streams and sustain operations during wide-area network outages.

    • Network (Telco) Edge: Embeds compute capacity inside cellular base stations, carrier aggregation sites, or Content Delivery Network (CDN) points of presence—often via Multi-access Edge Computing (MEC)—to serve nearby mobile and web clients over short network hops.

    • Managed Cloud Edge: Extends a cloud provider's APIs, security policies, and container orchestration onto provider-operated appliances or regional metro nodes, unifying deployment workflows across central clouds and remote locations.

    Example

    In an automotive assembly plant, high-speed cameras inspect welded body frames at 60 frames per second and stream video over a local gigabit switch to an on-premises edge server running a quantized defect-detection model. By analyzing frames locally in under 15 milliseconds, the edge node triggers the robotic line to halt defective parts immediately while transmitting only flagged images and hourly summary metrics to the central cloud for long-term model retraining.