Multi-Table Multi-Statement Transaction with Apache Iceberg

#icebergSummit 2025 breakout session delivered by Jack Ye, Iceberg Community member, and Drew Gallardo, engineer at #AWS. Session Description: This talk introduces a new open source storage-only spec that enables multi-table multi-statement transaction for #ApacheIceberg tables. In this talk, we will dive deep into: 1.Traditional database transaction concepts and its applications in a modern lakehouse 2.How this spec extends the Apache Iceberg transaction guarantees and implements ANSI-compliant transaction semantics across lakehouse objects with only storage level primitives 3.Design details of the spec using a b-tree data structure to ensure correctness, efficiency and scalability for concurrent transactions 4.Demo for running concurrent multi-statement transactions against multiple Apache Iceberg tables with SERIALIZABLE or SNAPSHOT ISOLATION guarantees using its SQL extension for Apache Spark 5.Demo of using it as a storage-only backend implementation for the Iceberg REST catalog server in Apache Gravitino 6.Next steps of the project to support features like more table formats (e.g. Apache Hudi, Apache Paimon, Lance), more objects (e.g. view, stream, index, policy), etc.