Building real-time analytics with ClickHouse and edge streaming

Real-time analytics at scale requires a fundamentally different architecture than traditional batch processing. When we set out to build the analytics platform for anielak.net, we knew we needed something that could handle millions of events per hour while maintaining sub-second query performance.

We chose ClickHouse as our analytics database for its columnar storage engine and vectorized query execution. On the ingestion side, we built an edge streaming layer using Kafka that collects events from our global Points of Presence and routes them to ClickHouse with minimal latency.

The result is a pipeline that processes over 10 million events per hour with p99 query latency under 200 milliseconds.