What is Edge Computing?
Updated: 3 days ago
Edge computing is a distributed computing approach that processes data close to the devices, users or systems that create and consume it, rather than sending every task to a distant central cloud or data centre. The important distinction is location: the compute workload moves nearer to the point where the data is generated. That can reduce network delay, reduce the amount of data crossing a broadband or wide area connection, and allow some applications to keep operating when upstream connectivity is limited or unavailable.
Edge computing does not replace cloud computing. In most real deployments the two work together. Time sensitive filtering, control or inference can happen locally at the edge, while central cloud platforms handle functions such as long term storage, fleet management, large scale analytics, model training and coordination across many sites.
The infographic below shows the basic relationship between devices, edge processing, cloud services and applications.
How does edge computing work?
A conventional cloud first design may send raw data across the network to a remote data centre, wait for processing, and then return a result. An edge design moves at least part of that processing closer to the source. The edge might be the device itself, a gateway in the same building, an on premises server, a small regional facility, or compute capacity integrated into a telecommunications network.
A typical flow is device or sensor → local network → edge compute → selected data or results → cloud platform. The exact boundary is not fixed. What counts as the edge depends on the application and where centralised infrastructure begins.
ETSI describes Multi access Edge Computing as providing cloud computing capabilities and an IT service environment at the edge of the network. ETSI also notes that edge deployments can range from on premises edge to network edge and are not limited to mobile access.
What happens at the edge?
Filtering: discard or aggregate data that does not need to travel upstream.
Analytics: calculate useful results locally from sensor, application or network data.
Control: trigger a local action without waiting for a remote service to respond.
Caching: keep frequently requested content or data closer to users.
AI inference: run a trained machine learning model locally while model training may remain centralised.
Protocol translation: allow an edge gateway to connect local industrial or IoT protocols with IP based cloud services.
Temporary storage: retain data locally during a connectivity interruption and synchronise it later.
Edge computing vs cloud computing
Cloud computing concentrates scalable compute and storage in data centres. Edge computing distributes selected compute and storage closer to endpoints. Neither architecture is automatically faster, cheaper or more secure in every situation. The right design depends on the workload, network path, amount of data, response time requirement, hardware capability and operational complexity.
Edge computing: prioritises proximity, local responsiveness and reduced upstream data movement.
Cloud computing: prioritises centralised scale, elastic resources, shared services and large scale storage or processing.
Hybrid edge and cloud: performs immediate or bandwidth intensive work locally while using the cloud for central management, aggregation and deeper processing.
For the centralised side of this relationship, see What is Cloud Platforms?. Edge workloads may also run inside virtual machines or containers depending on the platform.
Why can edge computing reduce latency?
Network latency is partly determined by how far packets must travel and how many systems they pass through. If an application can obtain a result from a nearby edge node rather than a remote data centre, the network portion of the response path can be shorter. That is particularly valuable for interactive control, industrial automation, computer vision, gaming and other workloads where waiting for a round trip to a distant service may be undesirable.
Edge computing cannot remove all delay. Processing time, WiFi or Ethernet behaviour, access network latency, queueing, application design and storage can still matter. See What is Latency? for the distinction between latency and connection speed.
Techie Broadband calculation: how much upstream data can local processing avoid?
Consider an illustrative camera stream averaging 8 Mbps. This is a theoretical example, not a claim about every camera or codec.
8 Mbps ÷ 8 = 1 MB/s.
1 MB/s × 86,400 seconds per day = 86,400 MB, or 86.4 GB per day using decimal units.
If an edge system analyses the video locally and only 5% of the original data needs to be sent upstream, the transmitted volume would be about 4.32 GB per day.
The avoided upstream transfer would be about 82.08 GB per day, a 95% reduction in this deliberately simplified example.
Real systems are more complicated. Video bit rate varies, metadata may be sent separately, important clips may need higher quality retention, and software updates or remote administration create additional traffic. The calculation demonstrates the architectural principle: processing and filtering locally can materially change how much data needs to cross the broadband connection.
That matters because bandwidth is finite. Edge computing can reduce demand on an upstream link, but it does not increase the physical capacity of that link.
Where can the edge physically be?
On the endpoint: a camera, vehicle, industrial controller, phone or other device performs its own processing.
On a local gateway: a router like appliance or dedicated gateway processes data from multiple nearby devices.
On premises: a server or small cluster in a factory, office, shop, hospital or other site runs local workloads.
At an access or network edge: a telecommunications or service provider places compute closer to subscribers than a conventional central cloud region.
At a regional edge: a smaller distributed facility serves a geographic area while remaining closer than a distant central region.
Edge computing and UK broadband
For a UK home or business, edge computing changes where application processing happens, not the underlying broadband access technology. FTTP, cable, fixed wireless or mobile broadband still carries traffic that needs to leave the premises. A local edge workload can continue to function without sending every byte upstream, but remote management, cloud synchronisation, software updates and external services may still depend on the internet connection.
Faster broadband can make central cloud services more practical by increasing available throughput, while lower latency can improve responsiveness. However, even a multi gigabit connection does not make distance, routing, queueing or remote application processing disappear. Conversely, an edge system does not make a slow or unreliable broadband connection faster. It reduces dependence on that connection for the parts of the workload that can genuinely be processed locally.
Common edge computing use cases
Industrial automation: analyse sensor data and respond to machinery locally.
Video analytics: detect events locally instead of continuously sending every video frame to a central service.
Retail: process local inventory, checkout or sensor data while synchronising selected information centrally.
Healthcare: process selected device data close to the point of care where architecture and governance permit.
Connected vehicles: perform time sensitive processing inside or close to the vehicle.
Energy and utilities: analyse data at remote assets where connectivity may be constrained.
Content delivery and applications: place selected compute or cached content geographically closer to users.
Edge AI: run inference close to the source while larger training and fleet management tasks remain centralised.
AWS describes edge computing as bringing storage and compute closer to devices and users, while Microsoft Azure's edge computing overview similarly emphasises processing data close to where it is created. These descriptions support the same architectural principle, although individual vendor products implement it differently.
What are the benefits of edge computing?
Lower network latency for workloads that can be served locally.
Lower upstream bandwidth consumption when raw data can be filtered, aggregated or discarded locally.
Greater local resilience when an application can continue operating during an upstream outage.
Faster local control loops for time sensitive systems.
Potential privacy and data governance advantages when unnecessary raw data does not leave the site.
Reduced dependence on central compute for tasks that are efficient to execute locally.
What are the limitations and trade offs?
More distributed hardware means more devices to patch, monitor and replace.
Physical security matters because edge hardware may sit outside a protected data centre.
Resource limits can be much tighter than in a large cloud region.
Local failures can affect a site even if the central cloud remains healthy.
Data consistency and synchronisation become more complex across many locations.
Security policy, identity, encryption and certificate management must work across a distributed fleet.
Some workloads are still better suited to centralised infrastructure because they need very large compute or storage resources.
Is edge computing the same as fog computing?
The terms overlap, but they are not always used identically. Edge computing broadly moves processing closer to endpoints. Fog computing is often used for an intermediate distributed layer between endpoint devices and a central cloud. In practice, vendors and researchers use the boundaries differently, so the architecture matters more than the label.
Is edge computing the same as a CDN?
No. A content delivery network is a specific distributed system for placing content and sometimes application logic closer to users. CDN infrastructure can be part of an edge strategy, but edge computing is a broader concept that also includes local gateways, on premises compute, telecommunications edge infrastructure and processing directly on devices.
How to decide whether a workload belongs at the edge
Identify the response time requirement. Determine whether a remote round trip is acceptable.
Measure how much raw data the workload generates and how much actually needs to leave the site.
Determine what must continue working if the broadband or WAN connection fails.
Check whether the edge hardware has enough CPU, memory, storage and acceleration for the workload.
Define what data can remain local and what must be retained or synchronised centrally.
Design security, patching, monitoring and identity for every distributed node.
Test the complete path under realistic load rather than assuming proximity alone guarantees performance.
How to troubleshoot an edge application
Confirm the endpoint or sensor is generating the expected data.
Check the local Ethernet or WiFi connection and negotiated link rate.
Verify that the edge service is running and has sufficient CPU, memory and storage.
Test local response time before involving the internet connection.
Check DNS, routing, firewall and certificate configuration for upstream services.
Measure broadband latency, packet loss and available bandwidth when cloud synchronisation is slow.
Compare local processing time with remote service time to locate the actual bottleneck.
Review logs on both the edge node and central platform before blaming the ISP.
Edge computing FAQs
Does edge computing need the internet?
Not necessarily for every local function. An edge system can be designed to process and act on data locally during an internet outage. Functions that depend on remote APIs, cloud storage, central management or synchronisation still require connectivity.
Is a home router an edge computing device?
A router sits at the edge of a local network, but simply routing packets does not automatically make it an edge computing platform. If it runs local applications, filtering, analytics or other compute workloads beyond ordinary forwarding, it can participate in an edge architecture.
Does edge computing make broadband faster?
No. It can reduce the amount of data that needs to cross the broadband connection and can shorten the path for locally processed tasks. It does not increase the line rate of the broadband service.
Is edge computing more secure than cloud computing?
Not automatically. Keeping some data local can reduce unnecessary exposure and data movement, but distributed edge nodes create their own security responsibilities. Physical access, patching, credentials, encryption, monitoring and fleet management all matter.
What is Multi access Edge Computing?
Multi access Edge Computing, or MEC, is the ETSI framework for providing cloud computing capabilities and an IT service environment at the network edge. Despite its telecom origins, ETSI notes that MEC can apply to mobile, fixed and WLAN access.
Key takeaway
Edge computing is about putting the right processing in the right place. It moves selected compute and storage closer to the devices or users that need them, which can reduce latency, reduce upstream data movement and improve local resilience. The cloud still remains valuable for central coordination, large scale storage and heavy processing. For broadband users, the practical point is that application performance depends not only on the speed of the internet connection but also on where the work happens across the complete end to end path.
To explore the related technical concepts used in this Edge Computing guide, visit our broadband terminology guide.
