Design a Web Analytics / Event Tracking Platform (Kafka + ClickHouse / Druid).

In Simple Words: Ingest, aggregate, and query 10 billion daily user interaction events (pageviews, clicks, scroll depth) with real-time dashboard visualization.

Storage Architecture: Columnar vs Row-Oriented:

• Row-oriented databases (PostgreSQL/MySQL) are designed for single-row OLTP operations.
• Columnar Databases (ClickHouse / Apache Druid) store table data column-by-column on disk. Reading and aggregating billions of rows (SELECT count(*), avg(duration) GROUP BY country) only reads the specific requested columns, compressing data by 90% and executing queries in milliseconds!

flowchart LR
    SDK["Browser / Mobile JS SDK"] -->|"HTTP POST (Batched)"| Collector["Event Ingestion Gateway"]
    Collector --> Kafka["Apache Kafka Cluster"]
    Kafka --> Flink["Apache Flink (Real-time Window Aggregation)"]
    Kafka --> ClickHouse["ClickHouse Columnar Storage"]
    ClickHouse --> Dashboards["Real-Time Analytics Dashboard UI"]