US Edge server meaning
An edge server is a server placed at the edge of a network, close to the machines, cameras or people that create and use the data, rather than in a central data centre or a cloud region. It brings processing, storage and networking on site, so data can be acted on in milliseconds and only the results need to travel.
The word “edge” is relative. For a manufacturer the edge is the factory floor, for a retailer the back office of each store, for a mobile operator the base station, and for a rail or bus operator the vehicle or depot. An edge server is whatever server hardware sits at that point and does work that would otherwise go to the data centre.
Edge servers sit in the middle of a three-tier model: small edge devices and gateways on each machine, edge servers on each site, and the cloud or central data centre for long-term storage and company-wide analysis.
On-premises and CDN edge servers
Search for the term and you will find two meanings. Both describe a server near the user, but they do different jobs and run on different hardware.
-
//
On-premises edge server. Installed on the customer’s own site to process local data: production lines, cameras, sensors, point-of-sale or building systems. This is the industrial meaning and the one this page covers.
-
//
Telecom or network edge server. Run by a mobile or network operator at a base station or local exchange, often called multi-access edge computing (MEC), so 5G applications get low latency.
-
//
CDN edge server. Run by a content delivery network in hundreds of points of presence. It caches copies of web pages, images and video from the origin server, so users download them from somewhere nearby.
-
//
Device edge. Not a server at all, but the compute inside a camera, controller or gateway. It handles single-device tasks and hands heavier work to an edge server.
Edge server vs data centre server vs edge device
An edge server does the same kind of work as a data centre server but lives in a very different place. The table shows where it differs from the servers above it and the devices below it.
| Aspect | Data centre server | Edge server | Edge device or gateway |
|---|---|---|---|
| Where it sits | Central data centre or cloud region | Factory, depot, store, cell site or vehicle, near the data | On the machine, camera or cabinet it serves |
| Typical job | Shared business and web workloads | Pools data from many sources; runs analytics, AI inference and virtual machines | Collects data from one machine or line and passes it on |
| Processor | Many-core Xeon or EPYC, often dual-socket | Many-core Intel Xeon, Core or AMD EPYC, often with GPUs | Low-power Intel Atom, Core or Arm, sometimes an NPU |
| Memory | Hundreds of GB, ECC | Tens to hundreds of GB, usually ECC | A few GB |
| Form factor | Full-depth rack in a cooled hall | Short-depth 1U–3U rack, tower or compact rugged chassis | DIN-rail box, mini PC or module |
| Management | BMC plus data-centre tooling | BMC for remote power, console and monitoring | Device management agent, sometimes none |
The gap between an edge device and an edge server is mostly scale. A single industrial IoT gateway reads one machine; an edge server takes the feeds from dozens of gateways, cameras and controllers and runs the heavier software on them.
What an edge server is used for
Edge servers earn their place when a workload is too big for a small device and too time-sensitive, too data-heavy or too private to send to the cloud.
-
//
Real-time analytics and control. Responding to sensor and machine data within milliseconds, for quality control, safety interlocks or robot guidance.
-
//
AI inference on many streams. Running vision models across tens of camera feeds with one or more GPUs, rather than a computer per camera. Our comparison of edge AI and cloud AI explains where each fits.
-
//
Data aggregation and filtering. Collecting raw data on site and sending only summaries or exceptions upstream, which cuts bandwidth and cloud storage costs.
-
//
Consolidation and virtualization. Running several applications, historians or virtual machines on one server instead of a row of separate PCs.
-
//
Local resilience. Keeping a site running when the wide-area link drops, then synchronizing with the cloud when it returns.
Edge server vs GPU server
A GPU server is defined by what is inside it, an edge server by where it goes. A GPU server is any server built around several full-size GPUs, with the power supplies, cooling and PCIe lanes to feed them, and most are designed for data centres: deep rack chassis, high airflow and power measured in kilowatts. An edge server may contain GPUs, but it has to suit the site first: shallow cabinets, noise and temperature limits in offices and plant rooms, and the power that is actually available.
The two meet when AI inference moves on site. Running vision or language models on many streams needs GPU capacity, but a full-depth multi-GPU chassis rarely fits a factory office or a depot comms room. Edge GPU servers close that gap with fewer GPUs in a tower or short-depth chassis. The Advantech SKY-602E3, for example, is a mid-tower GPU server for four double-width GPUs on an AMD EPYC Embedded 8004 processor, sized for a lab, office or edge room with no rack. Our edge AI servers cover the range between compact GPU systems and rack GPU servers.
What to look for in edge server hardware
The hardware has to deliver server features in a space that was never built for servers: a control cabinet, a store cupboard, a vehicle or a roadside enclosure.
Start with the space. Short-depth 1U and 2U chassis fit wall-mount and network racks, towers suit offices and stores, and sealed compact chassis go where there is no rack at all. Then check the environment: operating temperature, dust, vibration and whether the site can supply redundant AC or DC power. Inside, look for ECC memory, NVMe storage, PCIe slots sized for the GPUs or network cards you need, and a processor family with long availability, such as Intel Xeon or Core, or AMD EPYC Embedded. Our guide to AMD EPYC Embedded for edge servers covers one such platform.
Remote management matters more at the edge than in a data centre, because there is rarely anyone on site. A baseboard management controller gives out-of-band management: remote power control, console access and health monitoring even when the operating system is down.
Our rack mount servers include 1U, 2U and 3U edge AI servers on Intel Core, AMD Ryzen 9000 or EPYC 4005 processors, a fanless 2U rack server and a mid-tower GPU server on AMD EPYC Embedded 8004, with redundant AC or DC power on several models.
Quick answers on edge servers
Edge servers: common questions
What is an edge server used for?
Processing data where it is created. Typical jobs are real-time analytics, AI inference on camera feeds, filtering data before it goes to the cloud and running several site applications on one machine.
What is the difference between a CDN and an edge server?
A CDN is a network of edge servers. A content delivery network runs many edge servers in different locations to cache web content near users. An industrial edge server processes local data instead of serving web pages.
Is an edge server the same as edge computing?
No. Edge computing is the approach of processing data near its source. An edge server is one piece of hardware that does it, alongside edge devices and gateways.
What is the difference between an edge server and a local server?
Mostly the job and the setting. A local server usually sits in an office comms room and serves users. An edge server sits near machines or sensors, is built for that environment and feeds results to central systems.
Can an edge server work without the cloud?
Yes. It runs its applications locally and can keep a site working on its own. Most deployments still send summaries or models to and from a central system when a link is available.
What is the difference between an edge server and a GPU server?
Location versus contents. A GPU server is built around several GPUs, usually for a data centre. An edge server is built for the site it serves, and may or may not include GPUs.
We build edge servers around the site they will go into, and fit the processors, memory, storage, GPUs and operating system as standard. Tell us the workloads, the space and the power available, and we will specify the server to match.