To understand the meaning of the term “Data Lakehouse”, we need to first review the other storage options.

Data Warehouses became popular in the 1980’s as a means for businesses to store data for specific uses.  Data had to be structured and mapped for an intended purpose.

The Big Data of the early 2000’s, and the desire for AI and ML applications, made Data Warehouses difficult to manage.  Data Lakes were formed to house unstructured data alongside the structured data.  This unstructured data storage provided the opportunity for Data Scientists to tap into more ways to gain insights into the vast amounts of data organizations were generating, like images, videos, and more.  Unfortunately, the unstructured data led to complications which brought about the term “data swamps”.

Today’s data engineers and architects needed a hybrid approach to managing the data.  That is the idea behind the Data Lakehouse. Data Lakehouses apply metadata, cataloging, and indexing to unstructured data stored in cloud blob storage. This enables scaling compute and storage independently of each other, distributing data processing across a cluster, ACID compliant-transactions, support for SQL, Python, and Scala. The result being faster, more seamless accessibility for algorithms using computer vision, natural language processing, and other technologies.

Data Lakehouses are also more economical; incoming data is automatically integrated, fewer systems would need to be accessed, and data prep time is shorter.