Data lakehouses

A data lakehouse is a type of data product that serves as both a data lake and a data warehouse, with the idea that data lakes bring "the best of both worlds" together.

Typical data lake and data warehouse architecture treats each system independently, which increases the overall operational overhead and increases the overall data volume by requiring multiple copies of the same data. Data lakehouse architecture attempts to reduce that overhead by combining both architectures into a single system that can, when appropriate, share data in multiple contexts.

Data lakehouse organization

Data in a data lakehouse is typically organized into three tiers of data:

Bronze data is raw data, replicated from the source and sometimes decorated with metadata (such as the date each record was replicated). It is used for archival, reprocessing, and lineage tracing.
Silver data is bronze data, copied and transformed to create relationships and introduce surrogate keys between entities in a 3NF-like structure. It most typically is used for activities such as exploratory data analysis, ad-hoc reporting, and machine learning.
Gold data is silver data, copied and transformed into a star schema and optimized for read operations and analytical use. It is used primarily for formal reporting, dashboards, and self-service analytics.

Properties of an effective data lakehouse

Handles all types of data including structured, semi-structured, and unstructured
Provides the ability to explore, preview, and query data through both visualizations and code (e.g. SQL)
Reduces the need for ETLs through the ability to query the same data in multiple contexts
Manages meta data for searching, tracking, and sharing data.
Stores data in open data formats such as Parquet, JSON, and CSV.
Decouples storage from compute so that each can scale independently
Includes integrated security and governance controls that allow data to be properly classified and protected.
Serves all data functions within the organization.
Is cost effective compared to distinct data lakehouse and data warehouses solutions.