Data Lake
Aussi : Data Lake, Enterprise Data Lake
A data lake is a centralized repository that stores large volumes of raw data in its native format, from structured tables to unstructured files, until needed.
What It Is
A data lake is a centralized storage repository that holds vast amounts of data in its raw, native format. Unlike a traditional database or warehouse, it does not require you to define a schema before loading data. This is known as schema-on-read: structure is applied when the data is queried, not when it is stored.
A data lake can hold:
- Structured data (relational tables, CSV files)
- Semi-structured data (JSON, XML, log files, Parquet)
- Unstructured data (images, video, audio, free text, PDFs)
Why it matters
Organizations generate data faster than they can model it. A data lake lets teams capture everything cheaply now and decide how to use it later. This flexibility supports advanced analytics and machine learning, which often need raw signals that a pre-aggregated warehouse would have discarded.
Key benefits:
- Low cost storage using object stores (cloud or on-premises)
- Decoupled storage and compute, so you scale each independently
- Single source for analysts, data scientists, and reporting tools
The main risk is the data swamp: without governance, cataloging, and quality controls, a lake becomes an unsearchable dumping ground. Strong metadata, access policies, and a data catalog are essential.
How it is used in practice
Data lands in the lake through batch loads or streaming ingestion. It is commonly organized in zones: a raw zone (untouched source data), a cleansed zone (validated and deduplicated), and a curated zone (business ready datasets). Engines such as SQL query services, notebooks, and BI tools read directly from the lake.
Many teams now adopt the lakehouse pattern, which adds table formats and transactions on top of the lake to combine lake flexibility with warehouse reliability.
Concrete Example
A retailer streams clickstream logs, stores nightly sales tables, and dumps product images into one lake. Marketing queries clicks for campaign attribution, finance reads sales for reporting, and a data science team trains a recommendation model on clicks plus images, all from the same repository.