Deep dives on the Streaming Lakehouse, Apache Pulsar, BookKeeper, and building data systems that feed modern AI.
An end-to-end comparison of Kafka + Spark + Iceberg, Snowflake, ClickHouse, and Streaming Lake — measured read amplification (11–72× less I/O), a one-engine architecture with no object store, and a line-by-line cost model (~1.4× cheaper).
Read the comparison →Kafka, Iceberg, Fluss, Databricks, ClickHouse, Snowflake, Druid, Pinot, RisingWave — a global matrix, per‑technology tables, and why the real comparison is the whole pipeline.
Read comparison →Kafka + Iceberg, Apache Fluss, Hudi, RisingWave, Databricks — a capability matrix and the one thing none of them deliver.
Read comparison →One copy of data, three access patterns. How columnar pages inside BookKeeper unify messaging, AI features, and SQL.
Read deep dive →The segment-oriented BookKeeper architecture behind Pulsar — and why it's the right foundation for a lakehouse.
Read article →Why append-only streaming history gives you leakage-free, point-in-time-correct features — with no Iceberg snapshots.
Read more →