What is IoT and how does it work?

What is IoT and how does it work?

Connecting the physical world to useful information

Hello, friends. The Internet of Things describes systems in which physical objects exchange measurements, status information and, where needed, commands with software over a network. Examples include monitoring warehouse temperature remotely, collecting meter readings centrally or checking equipment status through an application. The aim is to use information for monitoring, decisions or control.

IoT is not the name of one device, wireless technology or app. Devices, connectivity, data processing, applications and operational processes work together. Data exchange can be automatic, while commissioning, permissions, maintenance and some decisions still involve people. AI, robots, phone apps and cloud services are not mandatory components of every deployment.

1. What do sensors, processors and actuators do?

A sensor detects a physical quantity, such as temperature or pressure, or a condition such as an object's presence. A processor handles the signal, stores it where necessary and prepares it for communication. An actuator, such as an electrically operated valve or motor, produces a physical effect. An IoT device may only measure, only control an output, or do both.

Every sensor does not need its own internet connection. Existing instruments can join a system through a PLC or data-acquisition unit. For example, a gateway can read a Modbus-connected meter and forward its information to higher-level software. This is a different integration route from installing a wireless module inside the original instrument.

2. How does information travel from the device to the user?

In a typical deployment, a device takes a measurement, sends it through a local connection to a gateway or directly to a server, and software interprets and displays it. A gateway is an intermediary that can transfer information between different communication environments. It may be a separate unit or a function integrated into a router, industrial computer or controller.

Bluetooth SIG's gateway material explains how Bluetooth devices can participate in internet-based applications through an intermediary. A Bluetooth connection alone does not provide internet access. The gateway also needs suitable connectivity, software and access permissions to reach the remote service.

3. Where does processing happen: on the device, locally or in the cloud?

With edge computing, processing happens near the data source, on a device or gateway. It can support local alarms, filtering and selected functions during connectivity loss. Cloud computing can provide remote storage, comparisons across facilities and user-access management. Both approaches can be used within the same system.

The W3C Web of Things architecture describes gateway and edge deployments rather than prescribing one arrangement for every IoT system. The cloud is optional, and continued operation during an internet outage is not automatic. Local behaviour must be designed and tested.

Editor's note: Assuming that every device must connect directly to the internet is a common mistake. A sensor may communicate only with a local gateway, which provides internet connectivity. Local functions may also continue without internet access. Conversely, simply labelling any isolated local-network device as IoT, without explaining further integration, provides little technical clarity.

4. Are Wi-Fi, Ethernet and MQTT the same kind of technology?

No. Ethernet, Wi-Fi, Bluetooth and cellular connections offer different forms of network access. MQTT, HTTP and CoAP are application protocols used to exchange information. MQTT messages can, for example, travel over a suitable Wi-Fi or Ethernet connection. Wi-Fi support does not establish compatibility with every application.

Choose connectivity according to distance, site conditions, power supply, data volume and acceptable latency. Frequent transmissions can increase battery use and traffic. For energy-saving applications, examine measurement intervals, sleep behaviour and reconnection requirements alongside the wireless technology. Streaming video and reporting a few readings per hour create different requirements.

5. What do MQTT and CoAP make easier?

MQTT uses a publish-and-subscribe model. A measurement device publishes information under a topic name to an MQTT server, commonly called a broker. Authorised subscribers receive the relevant messages. The broker can run locally or in the cloud, and every device need not know every other device individually. The OASIS MQTT standard defines this communication model and message-delivery behaviour.

CoAP is an application protocol designed for resource-constrained devices, offering requests and responses involving resources. IETF RFC 7252 describes its basic structure. Protocol choice depends on the device and application. Sharing a protocol does not automatically align field meanings, units or permissions: interoperability still needs verification.

6. What makes a measurement usable?

A value such as “23.4” is insufficient by itself. You need the device identity, measured quantity, unit, measurement time and data quality. Is it temperature or pressure? Which sensor produced it, and when? Arrival time at a gateway may differ from the actual measurement time. Clock synchronisation and time zones also affect comparisons.

Connectivity does not prove measurement accuracy. Sensor selection, installation, calibration and fault handling matter. Continuing to display the last valid reading after a sensor disconnects can mislead users unless the condition is clearly indicated. Online status and current, valid data should be treated as separate indicators.

7. Does message delivery prove that a command was executed?

Remote monitoring reads information; remote control can affect a physical process. MQTT quality-of-service levels concern message-delivery behaviour. A protocol acknowledgement does not prove that a valve opened or a motor turned. Suitable status feedback is needed to establish the physical result.

A control application needs defined permissions, command validity periods and duplicate handling. An old start command arriving after connectivity returns may no longer be appropriate. Receipt, acceptance and completion of a command can be tracked as separate states. Required local interlocks and safety functions need their own design.

8. A simple industrial monitoring example

In a hypothetical facility, a local controller reads tank level, a gateway forwards the information to a server, and an authorised operator sees it on a screen. The display provides level, measurement time and communication status. If communication fails, the last value is marked as stale. Whether local control continues is defined beforehand, rather than left dependent on the remote display.

This illustrates the difference between IoT and IIoT: availability, process conditions and the consequences of interruption become central in production environments. From an industrial cybersecurity perspective, a monitoring account should not receive unnecessary write permissions. Fast control loops and safety functions should not be made dependent on general internet connectivity without appropriate engineering.

9. Where do security and privacy begin?

IoT security starts with device identity and access permissions. Assess unique credentials, restricted interfaces, appropriate encrypted communication and verified update mechanisms together. NISTIR 8259A provides a framework for core device cybersecurity capabilities; it is not, by itself, a product certificate.

Encryption alone does not remove data-breach risks. Misuse of an authorised account or excessive retention can also cause problems. For privacy, establish what information is collected, why, who can access it and how long it is retained. Assess the need for continuous connectivity separately for each function.

10. What changes as the number of devices grows?

Connecting one device differs from operating a thousand. Plan for identity management at scale, software-version tracking, outages and data retention. As an illustration, 1,000 devices each sending 200 bytes of application data every minute generate 288 million bytes a day. That is approximately 288 MB per day, excluding protocol overhead, retransmissions, database indexes and backups.

Even with low average traffic, simultaneous reconnection can create a short burst of demand. Plan buffering during outages, retransmission and duplicate-record handling. Maintenance and updates and the end of vendor support are operational concerns too. Long-term costs extend beyond the device purchase price.

11. How should you start?

Define the actual requirement first: which information will improve which decision? Then use a small pilot to check measurement accuracy, outage behaviour, user permissions and data export. Success means obtaining a useful result from trustworthy information, beyond simply displaying readings. A manageable system with clear responsibilities is more valuable than collecting unnecessary detail.

Related guides

Fundamentals and applications: Main IoT guide, IoT in daily life and IoT in healthcare.

Installation and operation: Device compatibility, Maintenance and updates and Energy savings.

Trust and future developments: IoT device security, Privacy implications and The future of IoT.