In this article, we are looking into the MQTT packet format with relevant example packets. MQTT is a binary-based protocol and has command and command acknowledgement format. So every time a client sends a command to the broker, the broker sends an acknowledgement. This communication protocol is actually based on the TCP/IP protocol. So first there will be a TCP connection establishment and then there will be MQTT connection establishment and then the data transfer will occur. After which TCP connection will be terminated.

As this is a command and command acknowledgement based protocol, for each function the client needs to send commands to the broker. And they are sent as packets.
when a client has to publish a message following steps are made:

  • The client has to establish a connection to the broker by sending a connect packet (With username and password if needed)
  • Wait for the acknowledgement to see if the connection is accomplished or if there is an error.
  • The client will send a publish packet which will contain the topic name and message to be published.
  • wait for the publish response packet depending on the QoS level.
    QoS 0: There won’t be any response
    QoS 1: PUBACK – Publish acknowledgement response
    QoS 2 :

    • wait for PUBREC – Publish received.
    • send back PUBREL – Publish release.
    • wait for PUBCOMP – Publish complete.
  • If the communication is complete. The client can disconnect from the broker by sending a disconnect packet.

Similarly for subscribing to a topic :

  • Send a Connect packet to the broker (With username and password if needed)
  • Wait for the Connect acknowledgement packet from the broker to see if the connection is accomplished or if there is an error.
  • And when Connect acknowledgement is received, send the subscribe packet with the appropriate topic name.
  • Wait for the subscribe acknowledgement packet.

MQTT Packet Format

The MQTT packet consists of 2-byte fixed header + a variable header and a payload. In this first 2-byte fixed header will be always present in all the packets and the other two, variable header and payload are not always present.

MQTT Packet Format

Command type

Out of the two-byte fixed header, the first byte is the control field. This 8-bit control field Is divided into two 4 bit fields. The first 4 MSB bits are the command type field.
E.g. The value of the connect command is 1. That means for connect command the connect type field should be 1 which is 0001. For publish command the value is 3, therefore, the connect type field should be 0011.

mqtt header and mqtt command type

 

Control Flag bits

The next 4 bits are the control flag bits and they are used by the publish command for the rest of the commands they are reserved and the value will be 0.
For PUBLISH command:
0th bit denotes if the message that is published is retained the message.
1st and 2nd bits are used to select the quality of service if it is 0 or 1 or 2.
And the 3rd bit denotes if it is a duplicate message.

mqtt header and control flag bits

Remaining Length

The second byte of the fixed header contains the remaining length which is, length of variable header + the length of the payload. Remaining length can use up to 4 bytes in which each byte uses 7 bits for the length and the MSB bit being a continuation flag.
if the continuation flag bit of a byte is 1, it means the next byte is also part of the remaining length. And if the continuation flag bit is 0, it means that byte is the last one of the remaining length.

E.g. if the variable header length is 10 and the length of the payload is 20, the remaining length should be 30.

Variable header

A variable header is not present in all the MQTT packets. Some MQTT commands or messages use this field to provide additional information or flags and they vary depending on the packet type. A packet identifier is common in most of the packets types. We will discuss in detail the variable header for the CONNECT packet below.

Payload

In the end, the packet may contain a payload. Even the payload is optional and varies with the type of packet. This field usually
contains the data which is being sent.

E.g. For CONNECT packet the payload is client ID and ‘username and password’ if they are present. And for PUBLISH packet, it is the message to be published.

CONNECT packet

The first byte of the connect packet will be 10. Because the value of CONNECT command is 1, the first 4 MSB will be 1 and there are no flags so the next 4 bits will be 0.

mqtt connect packet

The second byte should be the remaining length. which is the length of the variable header and length of the payload. Let us decide this length after completing the variable header and payload.
You can see below the format of the variable header and payload for CONNECTpacket.

mqtt payload format

In the variable header, first, there should be the protocol name. And for this, the first 2 bytes should mention the length of the protocol name followed by the protocol name. In our case, the protocol name is MQTT which is of length 4. So it becomes :

MQTT CONNECT header

You cannot give whatever name you want here. As this name is used by the server to identify the MQTT traffic. If an invalid protocol is found, the server may reject the connection.

After the protocol name, there is the protocol level. This determines which version of MQTT it supports. For the version 3.1.1, the protocol is of level 4. And if the same protocol is not supported by the server it disconnects by sending an acknowledgement with return code 01.

After that, there is the connect flag byte. We can access the sandbox broker from Eclipse: https://iot.eclipse.org/getting-started/#sandboxes which does not require any username and password as it is open to the public. With a clean session and without a last-will message our CONNECT flag byte becomes 02.

MQTT CONNECT flag byte format

Next two bytes are used to mention the keep alive duration in seconds. For 60 seconds, the value will be 003C in hex.

After the variable header, there will be the payload and it will contain; client id, username and password. In our case there is no username and password, so only client Id will be present. Same as we did for protocol name, the first 2 bytes will denote the client id length. Let’s assume our client id is PQRST.

MQTT CONNECT packet example 1

Now it’s easy to determine the ‘remaining length’ which is supposed to be here. If you count the total bytes used for variable header and payload. It is 17. Therefore the ‘remaining length will’ be 17.

MQTT CONNECT packet example 2

CONNACK packet

Once the connect packet is sent, and if the broker receives the connection it will send back the acknowledgement – CONNACK. In the variable header of CONNACK there will be the connect return code. By reading that we can understand if the connection is established and if not, The reason behind rejection.

MQTT-CONNACK-packet-format

PUBLISH packet

Now let’s publish the message “HELLO” to the topic OPENLABPRO.
For the publish packet command value is 3. With QoS level 0 and without retaining the message control flag will be 0. In the variable header section first 2 bytes will denote the length of the topic and then followed by topic. In a similar way in payload section first 2 bytes will denote the length of the message which is followed by the message.

MQTT PUBLISH packet

*All numbers are in hex

SUBSCRIBE packet

Now the message is published and if there are any subscribers for that topic, They will receive the message. To subscribe to a topic the client has to send the SUBSCRIBE packet.
Command value of Subscribe packet is 8 and the Control flag is reserved and should be 2.
The variable header will contain a non-zero 16-bit packet ID. And as payload, there will be the topic to subscribe followed by requested QOS level.

To subscribe to the topic OPENLABPRO with QOS 0:

MQTT SUBSCRIBE packet

Now you have a clear picture of what data are sent in an MQTT protocol. And as you can see, all the commands and instructions are sent and received as bits. and that is why it is a binary-based protocol. Text fields such as username, password, topic etc are encoded as UTF-8 strings. It uses one to four, 8-bit blocks to represent a character. Here, As the username and password are sent as raw data, The security is less. So SSL certification is preferred to be a better option.

You can find the complete packets and description at http://docs.oasis-open.org/mqtt/mqtt/v3.1.1/os/mqtt-v3.1.1-os.html

Spread the love, share this