Edge Computing for IoT
This article explains how edge computing can be used in Internet of Things (IoT) systems to process data closer to where it is generated, and why this approach is useful for systems containing large numbers of connected devices.
I recommend to read following as the foundation for this article:
▶ On-Device vs Edge vs Cloud Computing
▶ Multi-Access Edge Computing (MEC)
1. Why Do We Need Edge Computing for IoT?
As discussed in the previous section, IoT systems can contain a large number of devices that continuously generate data. Sending all of this data directly to a remote cloud server is possible, but it is not always the most practical approach.
For example, consider a factory containing hundreds of sensors and cameras.
Sensors / Cameras
│
│ Data
▼
Network
│
▼
Remote Cloud
│
▼
Processing
**In this approach, all data generated by the IoT devices must be transmitted to the cloud for processing.
This can create several problems.
First, a large amount of data may need to be transferred over the network. A camera, for example, can generate much more data than a simple temperature sensor.
Second, some IoT applications require the data to be processed quickly. A factory machine may need an immediate response when an abnormal condition is detected. Sending the data to a remote cloud and waiting for the result may not be suitable for such applications.
Third, IoT devices may generate data continuously, even when most of that data is not important enough to be sent to the cloud.
For example, a camera may continuously capture video, but an application may only need to know when a particular object or event is detected.
Edge computing provides another option.
Sensors / Cameras
│
│ Data
▼
Edge Server
│
├── Process Data
├── Detect Events
└── Filter Data
│
▼
Cloud
Instead of sending all raw data to the cloud, an edge server can process the data close to the IoT devices.
For example, an edge server can analyze camera footage and send only detected events to the cloud.
Camera
│
│ Video
▼
Edge Server
│
│ "Defect detected"
▼
Cloud
The cloud can then be used for tasks such as long-term storage, historical analysis, centralized monitoring, and large-scale processing.
Therefore, edge computing allows an IoT system to divide its processing between the devices, edge infrastructure, and cloud.
The main idea of edge computing for IoT is to process data closer to the devices that generate it, rather than sending all IoT data directly to a remote cloud.
The edge server can perform tasks such as:
- Collecting data from multiple devices
- Filtering unnecessary data
- Combining data from different sensors
- Detecting events or abnormal conditions
- Performing local analysis
- Sending selected results to the cloud
2. IoT Edge Gateways
In many IoT systems, the IoT devices do not communicate directly with a general-purpose edge server. Instead, they communicate through an IoT gateway.
An IoT gateway is a device that sits between IoT devices and the wider network. It can collect data from multiple devices and provide communication between the local IoT devices and other computing systems.
For example:
IoT Devices
│
├── Sensor
├── Sensor
├── Camera
├── Machine
└── Robot
│
▼
IoT Edge Gateway
│
▼
Network
│
▼
Cloud
The gateway can use different communication technologies to communicate with IoT devices. For example, sensors may communicate with the gateway using protocols or technologies designed for low-power or local communication, while the gateway may use Ethernet, Wi-Fi, or a cellular network to communicate with the rest of the system.
An edge gateway can also perform some processing locally.
For example, it can:
- Collect data from multiple sensors
- Convert data between different protocols
- Filter unnecessary data
- Aggregate measurements
- Detect simple events
- Perform basic local analysis
- Forward selected data to the cloud
This makes the gateway more than just a network connection between the devices and the cloud.
For example, suppose a factory contains hundreds of temperature sensors. Instead of every sensor sending every measurement directly to the cloud, the gateway can collect the measurements and calculate useful information locally.
Temperature Sensors
│ │ │ │
▼ ▼ ▼ ▼
┌─────────────────┐
│ Edge Gateway │
│ │
│ Data Collection │
│ Filtering │
│ Aggregation │
└────────┬────────┘
│
│ Selected Data
▼
Cloud
The gateway can therefore serve as the first layer of computing and data processing outside the IoT devices themselves.
This is one of the most common ways in which edge computing is introduced into an IoT system.
3. Example: Edge Computing in a Smart Factory
Consider a smart factory containing hundreds of sensors, cameras, and industrial machines. These devices continuously generate data about the operation of the factory.
For example, sensors may measure:
- Temperature
- Pressure
- Vibration
- Machine speed
- Power consumption
Cameras may continuously monitor production lines to detect defective products or unsafe conditions.
A simple architecture could look like this:
Sensors / Cameras / Machines
│
▼
IoT Edge Gateway
│
▼
Edge Server
│
▼
Cloud
The IoT edge gateway collects data from the different devices and provides connectivity between the devices and the rest of the system.
The edge server can then perform more computationally intensive processing.
For example, the edge server may analyze camera images to detect defective products.
Production Camera
│
│ Video
▼
Edge Server
│
│ Image Analysis
▼
"Defect Detected"
│
▼
Factory System
Similarly, sensor data can be analyzed to detect abnormal machine behavior.
Machine Sensors
│
│ Sensor Data
▼
Edge Server
│
├── Normal Operation
│
└── Abnormal Condition
│
▼
Alert
Only the information that is useful for the larger system needs to be sent to the cloud.
For example, instead of continuously sending raw camera footage to the cloud, the edge system can send events such as:
"Defect detected on production line 2"
"Machine vibration exceeded threshold"
"Temperature exceeded safe limit"
The cloud can then store these events and historical data for longer-term analysis, reporting, monitoring, and other applications.
This example illustrates how an IoT system can divide its responsibilities across different layers:
IoT Devices
│
│ Collect Data
▼
Edge Gateway
│
│ Collect / Filter / Aggregate
▼
Edge Server
│
│ Analyze / Detect / Process
▼
Cloud
│
│ Store / Analyze / Monitor
▼
Applications
The important point is that not all IoT data needs to travel to the cloud. Data can be processed at different stages of the system, depending on what needs to be done with it.
4. What Processing Can Be Done at the Edge?
The edge does not have to process every piece of data generated by an IoT system. Instead, it can perform the processing that benefits from being close to the devices and forward the remaining data to the cloud.
For example, an edge system can perform data filtering.
Sensors
│
▼
Edge Server
│
├── Normal Data ──► Discard / Aggregate
│
└── Important Data ──► Cloud
A large number of sensor readings may not need to be stored or transmitted individually. The edge server can filter the data and forward only the measurements that are relevant.
The edge can also perform data aggregation.
For example, instead of sending individual temperature measurements from 100 sensors to the cloud, the edge server can calculate statistics locally.
100 Sensors
│
▼
Edge Server
│
├── Average Temperature
├── Maximum Temperature
├── Minimum Temperature
└── Abnormal Readings
│
▼
Cloud
The edge can also perform event detection.
For example, a camera may continuously capture video, but the application may only be interested when a particular event occurs.
Camera
│
│ Video
▼
Edge Server
│
├── No Event ──► Continue Monitoring
│
└── Event Detected
│
▼
Cloud
More computationally intensive processing can also be performed at the edge. For example, an edge server can run an AI model to detect objects in camera images or analyze sensor data to identify abnormal machine behavior.
Therefore, edge processing in an IoT system can include:
- Filtering
- Aggregation
- Event detection
- Data transformation
- Signal processing
- AI inference
- Anomaly detection
- Local decision-making
The result of this processing does not always need to be sent to the cloud. Some decisions can be made entirely at the edge.
For example, if an edge system detects that a machine is operating in an unsafe condition, it can immediately send a command to stop the machine.
Machine Sensors
│
▼
Edge Server
│
│ Unsafe Condition
▼
Stop Machine
The cloud can still receive the event and store it for later analysis, but the immediate decision does not have to depend on the cloud.
This is one of the most important uses of edge computing in IoT: the edge can turn raw device data into useful information or actions before that data is sent to the cloud.
5. Takeaway
Edge computing allows an IoT system to process and filter data closer to the devices that generate it. An edge gateway can collect, filter, aggregate, and prepare data from resource-constrained devices, while an edge server can perform more demanding processing such as AI inference, event detection, and anomaly detection.
The important idea is that IoT devices do not have to send all of their raw data to the cloud. The edge can filter unnecessary data, extract useful information, and process important events before forwarding selected data to the cloud. This can significantly reduce the amount of data that needs to be transmitted while allowing the IoT devices, edge infrastructure, and cloud to work together.
Edge computing gives IoT systems a nearby layer of computing and data filtering between the devices and the cloud, allowing data to be processed, filtered, and reduced closer to where it is generated.