Edge computing is running applications and processing data close to where the data is created or used, such as in a factory, a store, a hospital, a vehicle or a provider’s facility nearby, instead of sending everything to a central cloud or data center. Only results, summaries or selected data travel back to the center. Doing the work locally can cut response times, reduce the amount of data sent over the network and keep a site running when its connection to the cloud is lost.
At a glance
- Edge computing places compute near the source of data or the user, from a device or on-site server to a nearby regional facility.
- It typically complements the cloud rather than replacing it.
- Common drivers are fast response times, limited or costly bandwidth, operating through network outages and keeping certain data on site.
- The hard part is usually operations: deploying, patching, monitoring and securing equipment at many locations.
- “Edge” means different places to different vendors, so definitions vary.
What problem it solves
Central clouds are efficient, but they are far from where a lot of data is now produced. A factory line, a store full of cameras or a fleet of connected equipment can generate more data than is practical to send over the network, and some decisions, like stopping a machine or flagging a defect, need answers faster than a round trip to a distant region allows. If the internet link fails, a site that depends entirely on the cloud may stop working.
Edge computing moves the time-sensitive and data-heavy work closer. Local systems make decisions immediately, continue to run during outages and send only what is needed upstream. Some organizations also use edge processing to keep sensitive data, such as video, on site for privacy or regulatory reasons.
How it works
Where it runs. Edge can mean compute inside a device or gateway, a small server or rugged appliance at a site, a micro data center in a closet, or a provider’s facility in a nearby city or at a cell tower site. Cloud providers, telecom carriers and hardware vendors each sell products for different parts of this range.
What runs there. Workloads are often packaged as containers or virtual machines so the same application can run at every site. Typical examples include video analytics, machine monitoring, local control systems, point-of-sale and data filtering. More edge deployments now run AI models locally to classify images or detect anomalies.
Connecting back. Each site connects to the cloud or data center over the internet, SD-WAN, private networks, or wireless such as LTE, 5G or private cellular. The design decides what happens when that link fails, which data is buffered locally and how it is synchronized later.
Managing at scale. Because sites often have no IT staff, edge platforms rely on central management: remote provisioning, configuration, software updates and monitoring for every node. Physical security, encrypted storage and network segmentation matter more than in a locked data center. Our Edge Computing solutions page covers options, and our blog on edge computing security looks at the risks in more depth.
When it matters for buyers
- When sites must keep running during WAN outages. Retail, manufacturing, healthcare and logistics sites often can’t stop when the connection does.
- When data volume outgrows the network. Video and high-frequency sensor data are the usual examples.
- When response time is part of the requirement. Automation, safety and interactive applications may need answers locally.
- When rolling out IoT or AI at many locations. Plan the management, security and support model before buying hardware.
- When data must stay on site. Some privacy policies, customer contracts or regulations favor processing data locally.
Questions to ask vendors
- Which “edge” are you selling: on our premises, in your facility, or both?
- How are edge nodes deployed, updated and monitored centrally, and what happens during an update failure?
- How does the system behave when the connection to the cloud is lost, and how is data synchronized afterward?
- How is data at the edge encrypted, and how are devices protected if stolen?
- What hardware is supported, and who replaces failed units at remote sites, how quickly?
- Can we run our own containers or applications, or only yours?
- How is pricing structured: per node, per site, per device or per data volume?
How it differs from a CDN
A content delivery network (CDN) also puts servers close to users, which is why the two are often mentioned together. A CDN’s core job is delivery: it stores copies of content from your origin servers and serves them from nearby locations. Edge computing is about running applications and processing data near where that data is created, often on your own premises. Many CDN providers now let customers run code on their networks, which is one form of edge computing, but most on-premises edge deployments, such as a factory analyzing machine data, have nothing to do with a CDN.
Problems this helps with
- Sites that stop working when the internet connection drops
- Sending huge volumes of camera or sensor data to the cloud
- Applications that need faster responses than a distant cloud can give
- Managing and securing small servers at dozens of locations
