MQTT Poller: an endogenous framework for large-scale MQTT protocol performance evaluation

Abstract

MQTT is an application layer protocol that supports synchronous and asynchronous client communication. It employs an intermediary server/broker between clients and a topic system to send and receive messages. Thanks to its simplicity and low control cost, it is commonly used in the development of applications for the Internet of Things (IoT), and it is also of interest in several research works, such as those related to the scalability of the protocol. In this context, it is observed that there is a lack of performance evaluation and experiment orchestration instruments. This work presents a framework for evaluating the performance of the MQTT protocol in many scenarios. The development of the framework follows an endogenous communication model (i.e., based on the MQTT topic schema) for control and interaction between the architecture's elements. Starting from a reference performance evaluation tool (MQTTLoader), it is possible to configure an evaluation scenario that makes it feasible to run multiple instances of this tool, under the coordination of an orchestrator, on machines possibly hosted in different administrative domains. A case study demonstrates the functionalities and practicality of using the framework to conduct MQTT performance evaluation experiments.