Scalable Distributed Broker System for Very Large MQTT Networks

MQTT is a publish/subscribe protocol whose usage is growing a lot in recent years, especially in the field of Internet of Things (IoT) and Wireless Sensor Networks (WSN). The protocol is based on a central broker entity, which limits the number of publishers and subscribers and is considered a single point of failure. When numerous clients connect and subscriptions change frequently, the load and physical resource on the broker increases significantly. In this work, we propose a distributed broker system for MQTT, where multiple brokers are interconnected in a scalable fashion. The brokers collaboratively maintain two separate domains for exactmatch and wildcard subscriptions. We apply the Chord distributed hash table algorithm and the trie tree (radix tree) data structure in the exact-match and the wildcard domains, respectively, to maintain routing complexity and routing table size. The preliminary simulation results show that with relatively large number of brokers and subscribed topics, our proposed approach is more efficient in terms of forwarding overhead and routing table size compared with the standard broker bridging approach.