Data Lakehouse
A data architecture that combines lake-style flexible storage with warehouse-style management, reliability and analytical capabilities.
A lakehouse can reduce the cost and delay of copying data between a lake and a warehouse, letting BI, data science and AI teams work from one governed copy of the data.
A media company stores event data as Delta Lake tables in cloud storage, runs nightly Spark jobs to build curated tables, and serves the same tables to Power BI dashboards and a churn model, with access rules managed in one catalog.
A lakehouse still requires data modeling, governance and performance tuning. Table formats and engines differ in features, so mixing them can create compatibility issues, and some high-concurrency BI workloads still perform better on a dedicated warehouse.
A lakehouse keeps data in open file formats such as Parquet on object storage and adds a table format layer, such as Delta Lake, Apache Iceberg or Apache Hudi, that records metadata and transaction logs. This enables ACID transactions, schema enforcement, time travel and efficient updates. SQL engines and machine learning tools read the same tables, with a catalog providing governance.
Systems Architecture
Databricks Blog — What is a Data Lakehouse? — https://www.databricks.com/blog/what-is-data-lakehouse; CIDR 2021 — Lakehouse: A New Generation of Open Platforms that Unify Data Warehousing and Advanced Analytics — https://www.cidrdb.org/cidr2021/papers/cidr2021_paper17.pdf