Performance Evaluation of MQTT Broker Servers

Internet of Things (IoT) is a rapidly growing research field, which has enormous potential to enrich our lives for a smarter and better world. Significant improvements in telemetry technology make it possible to quickly connect things (i.e. different smart devices) situated at different geographical locations. Telemetry technology helps to monitor and measure the devices from remote locations, making them even more useful and productive at a low cost of management. MQTT (MQ Telemetry Transport) is a lightweight messaging protocol that meets today's smarter communication needs. The protocol is used for machine-to-machine communication and plays a pivotal role in IoT. In case the network bandwidth is low, or a network has high latency, and for devices having limited processing capabilities and memory, MQTT is able to distribute telemetry information using a publish/subscribe communication pattern. It enables IoT devices to send or publish information on a topic head to a server (i.e. MQTT broker), then it sends the information out to those clients that have previously subscribed to that topic. This paper puts several publicly available brokers and locally deployed brokers into experiment and compares their performance by subscription throughput i.e., in how much time a broker pushes a data packet to the client (the subscriber) or how much time a data packet takes to reach the client (the subscriber) from the broker. MQTT brokers based on the latest MQTT v3.1.1 version were evaluated. The paper also includes mqtt-stresser and mqtt-bench stress test results of both locally and publicly deployed brokers.