Cloud object stores have enticed technologists for over a decade with the promise of cost-effective storage, enabled by the separation of storage and compute. However, data lakes built on cloud storage have not provided the same performance and consistency benefits as DBMS built on distributed file systems or custom storage engines. To solve this problem, Databricks developed and OSS’ed Delta Lake, an ACID table storage layer over cloud object stores. The design of Delta Lake – which Databricks describes as a “lakehouse” – hinges on a transaction log, compacted into Apache Parquet format and stored in the cloud object store. This log maintains information about which objects are part of a Delta table in an ACID manner and enables time travel and faster metadata operations. With Delta Lake, users can still separately scale compute and storage without running servers to maintain state. Therefore, the authors posit that Delta tables can replace previously siloed data lake, data warehouse, and streaming storage systems.