Adaptive middleware for costly data generation over MQTT

Message Queue Telemetry Transport (MQTT) is a universally adopted protocol for device-to-device communication based on a publisher/subscriber model. However, producers employing MQTT always produce and publish data regardless of subscription. This causes a waste of processing and communication resources when there is no subscriber, especially when the data generating cost is high. This work proposes the design of an MQTT middleware that facilitates exchange of heavy producing cost data in order to avoid unnecessary load when there is no interest in the content. The middleware operates between the application and the unmodified MQTT layer. The experiment was conducted on a real test bed to evaluate the efficiency of the proposed middleware compared to the original MQTT in scenarios with different numbers of subscribers. The evaluation indicates that the proposed method significantly reduces unnecessary traffic when there is no subscriber, while maintaining a small overhead ratio when the number of subscribers goes up.