What Is a Data Lake? Definition, Architecture & Examples

Data Lake Definition: A data lake is a centralized storage repository that holds vast quantities of raw data in its native format — structured, semi-structured, and unstructured — at any scale. Rather than requiring predefined schemas prior to loading, a data lake applies schema-on-read, allowing analytical engines and machine learning workloads to parse and transform data on demand.
Modern organizations generate unprecedented volumes of operational data across relational databases, web clickstreams, IoT telemetry, and documents. A modern cloud data lake stores all these raw assets securely in low-cost cloud object stores (like AWS S3, Google Cloud Storage, or Azure Blob Storage) and pairs them with modern open table formats like Apache Iceberg, Delta Lake, or Apache Hudi to enable ACID transactions, time travel, and ultra-fast SQL analytics.

Data Lake Definition: How Schema-on-Read Works

To truly define a data lake, you must understand the distinction between schema-on-write and schema-on-read:
In a traditional relational database or data warehouse, data must conform to a rigid, predetermined schema before it can be written to disk (schema-on-write). If an incoming record has an unexpected column, mismatched data type, or unparsed JSON payload, the ingestion job fails or rejects the record.
In contrast, a data lake adopts schema-on-read. Incoming data is written directly to disk in its raw, unaltered state. Whether you are streaming CDC events from RDBMS databases like MySQL or PostgreSQL, ingesting web logs, or uploading image files, data is persisted immediately without transformation bottlenecks. Schemas are projected only when querying or training ML models, allowing multiple teams to interpret the exact same underlying raw data in completely different ways.

What Type of Data is Stored in a Data Lake?

One of the defining benefits of a data lake is its ability to ingest and retain all three primary data structures without discrimination:
Data StructureCharacteristics & FormatsCommon Real-World Examples
Structured DataHighly organized tabular rows and columns with strict data types and relational constraints.Snapshots and CDC streams from PostgreSQL, MySQL, Oracle, SAP, Salesforce, and financial ledgers.
Semi-Structured DataData that contains self-describing tags, keys, or nested hierarchies without a fixed relational schema.JSON API responses, Kafka payloads, Avro files, Parquet columnar data, XML feeds, and web server access logs.
Unstructured DataData with no predefined conceptual model or formatting, making up over 80% of modern enterprise information.PDF reports, customer emails, audio recordings, video streams, medical imaging (DICOM), and vector embeddings.

Data Lake Architecture: The Medallion Multi-Hop Pattern

Without structured architecture, an unmanaged data lake can rapidly degrade into an unusable "data swamp". Leading enterprises adopt the Medallion Architecture (Bronze, Silver, Gold pattern) to govern and refine data as it flows through the platform:
  • Bronze Layer (Raw Ingestion Zone): Exact, immutable replicas of source data stored in native formats (JSON, CSV, Avro, or raw database dumps). Retains full historical lineage for replayability and regulatory audits.
  • Silver Layer (Cleansed & Enriched Zone): Filtered, deduplicated, and standardized datasets converted into high-performance columnar formats (Parquet or ORC) wrapped in open table formats like Apache Iceberg or Delta Lake.
  • Gold Layer (Curated Analytics Zone): Business-level dimensional aggregates, star schemas, and feature tables optimized for direct consumption by BI dashboards, executive reporting, and predictive AI models.
Open Table Formats: In modern data lakes, file formats (Parquet) are governed by table format specifications such as Apache Iceberg, Delta Lake, and Apache Hudi. These formats bring ACID transactional guarantees, hidden partitioning, atomic upserts/deletes, snapshot isolation, and time-travel querying to raw cloud object storage.

What is a Data Lake Used For? Key Enterprise Use Cases

Enterprises deploy modern data lakes to unlock versatile analytics workloads that traditional relational systems cannot handle:
  • Generative AI & LLM Training: Data lakes act as the foundational training store for LLMs, holding multi-terabyte corpora of unstructured PDFs, chat logs, transcripts, and high-dimensional vector embeddings.
  • Change Data Capture (CDC) & Streaming: Continuous ingestion of live operational database events from MySQL, PostgreSQL, and Oracle via Kafka and Debezium, keeping downstream data lakehouse tables synchronized in near real-time.
  • Ad-Hoc Interactive SQL Analytics: Using serverless query engines such as Presto, Trino, Google BigQuery Omni, or AWS Athena, analysts can run complex SQL queries directly against raw files in object storage without paying data warehouse ingest penalties.
  • Cost-Effective Long-Term Archival: Object storage lifecycle policies automatically transition cold historical datasets into coldline, glacier, or archive tiers, preserving institutional knowledge at pennies per gigabyte.

Is Hadoop a Data Lake or a Data Warehouse?

A frequent point of confusion is whether Hadoop is considered a data lake or a data warehouse. Hadoop is the pioneer of the data lake concept, not a data warehouse.
In the early 2010s, the Apache Hadoop ecosystem — centered around the Hadoop Distributed File System (HDFS) and MapReduce/YARN — introduced the world to storing unstructured and semi-structured data on commodity server hardware. It was the first widely adopted "Data Lake 1.0" implementation.
However, on-premises Hadoop clusters suffered from coupled storage and compute, high operational maintenance overhead, and complex hardware management. Today, Modern Cloud Data Lakes (Data Lake 2.0 / Lakehouses) have largely superseded on-premise Hadoop by leveraging elastic cloud object storage (Amazon S3, Google Cloud Storage, Azure Blob Storage) with serverless Spark compute (Google Cloud Dataproc Serverless or AWS EMR) and Apache Iceberg open table formats.

Data Lake on the Cloud: A Modern Foundation for Your Data - AWS, GCP & Azure

What is a Data Lake Architecture Overview
Full support of reading various sources like structured, semi-structured, and unstructured data and ingestion into a centralized Data Lake in cloud storage such as AWS S3, GCP GCS, and Azure Blob Storage using open table formats such as Apache Hudi, Apache Iceberg, or Delta Lake.
Our team of certified data engineers and architects specializes in implementing production-grade data lakes on cloud platforms like AWS, GCP, and Azure. We provide end-to-end solutions, from data ingestion to processing, storage, and cataloging.
Key architectural components delivered in every cloud data lake project:
  • Decoupled storage and compute architecture design (S3/GCS/ADLS)
  • Open table format integration (Apache Iceberg, Delta Lake, or Apache Hudi)
  • Enterprise Data Catalog setup (AWS Glue, GCP Dataplex / Data Catalog, Azure Purview)
  • Cloud cost governance and automated lifecycle archival policies
  • Granular IAM role-based access control (RBAC) and column-level masking
  • Automated data quality assertions and schema evolution management
  • DataOps telemetry, pipeline health alerting, and monitoring

Frequently Asked Questions About Data Lakes

What is a Data Lake?

A Data Lake is a centralized storage repository that holds vast amounts of raw data in its native format until needed. It stores structured (relational databases), semi-structured (JSON, Parquet, XML), and unstructured data (text, logs, media) at massive scale, applying schema-on-read during query execution rather than during ingestion.

What is the definition of a data lake?

By definition, a data lake is an enterprise data storage architecture that ingests all forms of raw organizational data into low-cost object storage without requiring upfront schema transformations. Consumers can then transform and query subsets of that data on-demand for machine learning, BI reporting, and predictive analytics.

What type of data is stored in a data lake?

A data lake stores three primary categories of data: (1) Structured data from relational databases (PostgreSQL, MySQL, Oracle); (2) Semi-structured data such as JSON API payloads, web server logs, CSV, and Parquet files; and (3) Unstructured data including audio, video, images, PDFs, sensor streams, and social media feeds.

What is a data lake used for?

A data lake is used for machine learning and AI training, business intelligence reporting, ad-hoc exploratory analytics, streaming IoT telemetry ingestion, and long-term cost-effective data archival. It provides a single source of truth across siloed business departments.

Is Hadoop considered a data lake or a data warehouse?

Hadoop (specifically HDFS) was the pioneer of the original on-premises data lake architecture. Hadoop is not a data warehouse. Modern architectures have largely replaced on-premises Hadoop HDFS with cloud object storage (Amazon S3, Google Cloud Storage, Azure Blob Storage) coupled with modern open table formats like Apache Iceberg.

What is the difference between a data lake and a data warehouse?

A data lake stores raw data in any format (structured, semi-structured, unstructured) and applies schema on read, making it flexible for data exploration, ML, and streaming. A data warehouse stores processed, structured data with a fixed schema applied on write, optimised for BI reporting and SQL analytics. Read our detailed comparison in Data Lake vs Data Warehouse.

What open table formats are used in a modern data lake?

The three industry-standard open table formats are Apache Iceberg (hidden partitioning, time travel, broad engine support), Delta Lake (ACID transactions and Spark ecosystem integration), and Apache Hudi (fast CDC upserts with Copy-on-Write and Merge-on-Read table types).

Need a Cloud Data Lake solution implementation / Bootstrapping in your organization?


Visit our product catalog for ready-to-deploy ingestion templates and our contact us page for a free consultation on enterprise data lake design.