Scaling MQTT With Apache Kafka

This document discusses scaling MQTT with Kafka to support more than 2 million connected publishers sending over 65,000 messages per second to a single subscriber. It outlines problems with directly scaling MQTT, including load balancing brokers and handling a single subscriber. The document proposes using Kafka as a backend for MQTT brokers to allow for horizontal scaling, load balancing of subscribers, and guaranteed delivery. Results were linear scaling for high throughput to a single subscriber. Remaining areas are security and configuration. The overall discussion is about using Kafka to enable MQTT to meet large IoT deployments of millions of devices and high message rates.