Kafka Architecture and Its Limits
Jagadeesh Rajah architects Kafka at Fortum, the Finnish energy company spanning B2B, B2C, and recycling. He came to it from an integration background, after Azure integration proved too slow for what his users needed. That experience shapes how he thinks about where one platform ends and another begins.
In this conversation
- →Why Fortum moved to Confluent Kafka
- →Where Kafka's future sits with AI, MCP, and context
- →What good Kafka architecture looks like
- →His hot take: no single platform, Confluent included, can solve everything
- →Where API management, throttling, and a listening layer take over
"Confluent cannot solve everything. For example, throttling. If you want to avoid certain users bombarding a topic, you should have throttling with API management."
Interview highlight
One platform can't do everything
Throttling, traffic shaping, knowing your consumers. Those belong to an API-management layer, not the streaming platform.