A quick introduction to MQTT message broking and the challenges of clustering



A publish/subscribe messaging protocol called MQTT (Message Queuing Telemetry Transport) runs on top of the TCP/IP protocol. In 1999, Andy Stanford-Clark of IBM and Arlen Nipper of Cirrus Link created the initial iteration of the protocol. MQTT messages can be as tiny as 2 bytes, making it faster than sending HTTP requests with your IoT device.

When a server receives information from one client using MQTT, it instantly sends that information to all other clients who have expressed an interest. HTTP requires headers containing a lot of information that other devices might not care about. Additionally, you must send a POST action to each client if you use HTTP and have many devices waiting for a request.

Although devices and Azure IoT Hub endpoints can connect via a port, Azure IoT Hub is not a fully functional MQTT broker and does not adhere to all MQTT v3.1.1 standard behaviors. The Azure IoT Hub uses supported MQTT behavior to boost its effectiveness.


It would be helpful to comprehend some of the terminology used and how each component fits together to make your network before you learn how to establish an MQTT network.

Broker – The server that sends information to connected customers who are interested is known as a broker.

Client – The computer or device connects to the broker to send or receive data.

The message’s topic is its subject. Customers can publish a topic, subscribe to it, or do both.

Publish Clients who provide information to the broker for distribution to customers who express interest in the issue.

Subscribe – Customers inform the broker of the topic(s) they are interested in. Any message sent to the broker by a client subscribing to a case is sent to all of the client’s subscribers.

Quality of Service, or QoS: Each connection can inform the broker of the quality of service using an integer value between 0 and 2. Only amongst MQTT clients does the QoS impact how TCP data streams are handled.

  1. Denotes a maximum of once, or once and only once, without requiring a receipt for the delivery. This practice is known as “fire and forget.”
  2. Makes at least one specification. The message is transmitted several times until acknowledged delivery—another term for accepted delivery—is made.
  3. The second specifies only once. A two-level handshake between the sender and receiver clients, or ensured delivery, guarantees that only one copy of the message is received.


Who Uses MQTT?

MQTT was initially created for the high-latency, low-bandwidth data lines used in the oil and gas sector. MQTT, however, is currently utilized in a wide range of applications outside the oil and gas industry, including managing intelligent lighting systems, Facebook Messenger, and AWS IoT. Overall, MQTT seems to be the protocol that works best for industrial enterprises’ control systems, and we can anticipate that its current rate of adoption will only continue to increase.

Challenges of Clustering

An illustration of the difficulties with clustering in MQTT. A distributed system that simulates a single logical MQTT broker is known as an MQTT broker cluster. It comprises numerous distinct MQTT broker nodes, typically deployed on various physical machines and connected via a network. A group of brokers functions like a single MQTT broker to an MQTT client.

The only example accessible to everyone is a temperature sensor in the home. A single MQTT broker, such as EMQX edge edition running on a Raspberry PI, should be more than sufficient to support intelligent home devices. An EMQX node can also support up to 2 million connections.

Imagine there were millions of cars on the planet, millions of streetlights across the nation, and so on. The sheer volume of devices (MQTT clients) and data would be so great that it would be beyond the capacity of any one MQTT broker.

This is among the justifications for building a cluster of MQTT brokers. But it also brings about new difficulties, such as:

  • How should clients determine which MQTT broker endpoint to connect to? That a client disconnects from one node and reconnects to another, MQTT subscriber session takeover;
  • All nodes in the cluster must constantly share the global routing table.
  • The first two problems can be effectively addressed by placing a load balancer in front of the cluster.

The TCP/IP protocol is built on top of the MQTT (Message Queuing Telemetry Transport) protocol. Using MQTT, a server quickly delivers information to all other clients who have expressed interest after receiving it from one client. Customers can subscribe to a topic, publish it, or do both. MQTT was primarily developed to work with the low-bandwidth, high-latency data lines employed in the oil and gas industry. MQTT is used for various purposes, such as controlling Facebook Messenger, AWS IoT, and intelligent lighting systems. Clustering creates new challenges, such as how customers choose which broker endpoint to connect to.

You may create valuable Azure IoT Hub goods and services with the self-service IoT platform akenza.io. It connects, manages, and keeps track of Internet of Things devices in a single location. Akenza.io is convinced that enterprises will experience a significant decrease in workload and complexity by implementing IoT technologies.