We revisit our previous benchmark, where Kafkorama scaled vertically on a single node to deliver 1 million messages per second to 1 million WebSocket clients in real time (with a median latency of 3 ms). In this new benchmark, we replace the single Apache Kafka® broker used previously with a production-ready Kafka service provided by Confluent Cloud. The results show nearly identical performance, with end-to-end latency increasing by only 2 ms at the median and 3 ms at the 75th percentile, while latency stability remains unchanged.

Kafkorama exposes real-time data from Apache Kafka® as Streaming APIs, enabling any developer — not just Kafka developers — to go beyond backend apps and build real-time web, mobile, and IoT apps for Kafka. For this to work at Internet scale, Kafkorama must be both fast and scalable to handle all your users and Kafka streams. In this post, we present benchmark results showing how Kafkorama scales both vertically on a single node and horizontally across a multi-node cluster — delivering 1 million messages per second to 1 million concurrent clients over WebSockets with end-to-end median latency of 3 ms.

We're preparing something powerful: a real-time API management platform for Apache Kafka®, designed to scale to millions.