In general, MQTT works according to the client-server principle. The server represents the so-called MQTT broker, which is responsible for message management, message distribution and storage. In MQTT, the clients are referred to as publishers and subscribers. The messages are either sent or received by the clients. The publishers take on the task of publishing and actively send messages to the broker. Subscribers, on the other hand, subscribe to certain information and receive the data relevant to them from the broker. This means that senders and receivers are independent of each other and cannot communicate directly with each other, so there is no end-to-end connection.

The message format in MQTT is structured in so-called topics.
The principle of these topics is structured as follows, for example: House/Floor/Room/Temperature.
This allows subscribers to subscribe to different variants of these topics in order to receive the required data from the MQTT broker. For this purpose, data from houses, floors and rooms can be subscribed to the broker individually or by inserting a wildcard "#".
MQTT uses TCP as the transport protocol for this data transfer. Ports 1883 and 8883 have been reserved for MQTT by the IANA(InternetAssignedNumbersAuthority).

In the event of a connection failure, it is possible to specify in advance whether and how data transmission should be continued after recovery. In this scenario, you can choose between three types of Quality of Service (QoS0, QoS1, QoS2). QoS0 stands for at-most-once delivery and is used when only a low transmission guarantee is required. Although an attempt is made to deliver the message, it either reaches the recipient once or not at all. QoS1, on the other hand, stands for at-least-once delivery. Here a confirmation from the recipient is expected; if this is not received, the message is resent. QoS2, Exactly-once Delivery, guarantees the transmission of a message exactly once.