AI glossary · Basic · Data engineering
Data lake
A data lake is a large-capacity repository that cheaply stores all kinds of data in one place in its original form, from tabular data to logs, images and documents. Instead of defining a format in advance, data is stored first and read in whatever way is needed at the time of use.
About this term
- Level
- Basic · Technical terms you often meet in AI news
- Category
- Data, ops & platforms
- Fields
- Data engineering
- Added
- 2026-10-09
How it works
A data lake is usually built on cheap, almost infinitely scalable space such as cloud object storage. Database tables, server logs, sensor records, PDFs, images and audio files are stored as files in their original formats. No structure is enforced at write time; instead, the structure needed is applied when the data is read for analysis. In practice, it is often managed in stages, divided into zones such as one for raw data, one for cleaned data and one for data ready for analysis.
Why it matters
AI training needs large amounts of unstructured data such as text, images and audio, not just tabular data. A data warehouse is strong with organized tabular data, but storing such varied raw data as-is in one is expensive and inconvenient. A data lake lets you 'keep it for now even if you don't know how it will be used later,' so when a new analysis or model training becomes necessary, you can start again from the raw data. That is why it is often used together with ELT-style data pipelines.
Data lake and warehouse
Data lakeStored in original form
- Anything: text, images, logs, etc.
- No structure enforced at write time
- Cheap, large-scale storage
- Without management, a 'data swamp'
Data warehouseOrganized tabular data
- Strong at structured data analysis
- Stored in a predefined structure
- Inconvenient for storing unstructured raw data
공통 The architecture that combines the strengths of both is the data lakehouse
What to keep in mind
If data just piles up without any rules, it becomes a 'data swamp' where no one knows what is where or whether it can be trusted. To prevent this, an inventory must be maintained with a data catalog, and access rights and quality rules must be set through data governance. Also, because data is stored as files, it has had the weakness that updates and deletions are hard to handle safely the way a traditional database does. The architecture that addresses this weakness by combining the strengths of a warehouse is the data lakehouse.
Example
Consider a video service that stores viewing logs, subtitle files, thumbnail images and customer inquiry records all in a data lake in their original form. The analytics team pulls viewing logs from it to produce usage statistics, while the AI team pulls subtitles and inquiry records from the same repository to use as training data for a recommendation model or a customer support chatbot.