Scheduling messages within MQTT shared subscription group in the clustered cloud architecture

摘要

The increasing number of objects are being connected to the Internet at an incredible rate, forming the Internet of things (IoT). With this volume of new devices on the Internet, service providers must assure to provide a fault-tolerant system that will manage to serve all its clients. Clustering is considered to be the most successful method for resolving this problem, as it ensures high availability. On the other hand, mechanisms which allow cluster components to distribute the work load, and use the available resources optimally, are also needed. In IoT applications, MQTT protocol (Message Queuing Telemetry Transport) is commonly used to connect end devices with the cloud services. It supports shared subscriptions, which allow load distribution among the subscribers to a given topic. In this paper, we propose the algorithm for MQTT message routing within the MQTT broker cluster, which we refer to as resource-levelling forwarding (RLF). We compare this approach with the already implemented forwarding methods such as round-robin and sticky algorithms.