Choose the Pattern, Not the Tool
4 entriesData architecture is less about choosing the latest tool than about choosing the right pattern for your business, data and scale. Each pattern below solves a different problem, and real-world data platforms rarely rely on just one: they combine several according to scalability, governance, latency and business needs. For data engineers, knowing when and why to use a pattern matters as much as knowing the tools behind it.
Refining and Processing Data
5 entriesMedallion organises data into progressively refined layers. Lambda and Kappa differ in how they serve batch and real-time needs: Lambda runs a batch layer and a speed layer side by side, while Kappa treats everything as a stream.
Lakes, Warehouses and Lakehouses
10 entriesA data lake offers flexible, low-cost storage for data in raw and varied formats. A data warehouse holds structured, modelled data for analytics. The lakehouse aims to combine the lake's flexibility with the warehouse's reliability, using open table formats and transactional metadata on object storage.
Ownership, Governance and Integration
8 entriesData mesh decentralises ownership to business domains, treating data as a product with federated governance and a self-serve platform. Data fabric uses metadata, governance and integration to connect data across sources, usually through a unified catalog and automation. They address different problems and are sometimes combined.
Modelling for History and Auditability
3 entriesData Vault is a modelling method built for history, auditability and scalability, separating business keys (hubs), relationships (links) and descriptive history (satellites).
Event-Driven and Streaming Architectures
9 entriesEvent-driven systems communicate asynchronously through events. Modern streaming architectures carry real-time data from source to analytics, usually with a broker such as Kafka, a stream processor such as Flink and change data capture from operational databases.
Related Pages
4 entriesNeighbouring pages on this site.
Frequently Asked Questions
What is the difference between Lambda and Kappa architecture?
Lambda keeps a batch layer for accurate, reprocessed results and a speed layer for low-latency results, which means two code paths to maintain. Kappa uses a single streaming path and reprocesses by replaying the log.
Is a lakehouse the same as a medallion architecture?
No. A lakehouse is a platform pattern that combines lake storage with warehouse-style management; medallion is a way of organising data in layers and is commonly used inside a lakehouse.
Are data mesh and data fabric alternatives?
They address different problems: data mesh is mainly about organisational ownership of data products, while data fabric is mainly about metadata-driven integration and governance. Some organisations use both.
When is Data Vault a good fit?
When you need full history, traceability and the ability to add sources without reworking the model, typically in the raw or integration layer of a warehouse.
Can I combine patterns?
Yes, and most platforms do, for example medallion layers on a lakehouse fed by streaming ingestion and governed through a data mesh operating model.
How current is this page?
Links were checked in October 2026. The list of patterns and their one-line summaries follow a short note on data architecture patterns supplied for this page; the explanations and sources here come from each pattern's primary or vendor-neutral references where available. Patterns overlap and vendors use some names differently, so check definitions before relying on them. Sites that block automated checks, such as lambda-architecture.net and the O'Reilly Radar, are kept and may need a manual click. Listing is not an endorsement.