Home  Data Services  Data Lake Architecture

Data Lake Architecture
— lakehouse, open formats

Lakehouse foundations on open table formats — medallion layers, unified governance and domain-oriented data products — on Microsoft Fabric, Databricks, Snowflake or native hyperscaler services.

Talk to us about this All Data services

Data Lake Architecture

DATA SERVICES

Topics we cover

Data lake architectureData lakehouseMedallion architectureOpen table formatsApache IcebergDelta LakeApache HudiMicrosoft FabricOneLakeDatabricksSnowflakeUnity CatalogZero-copy / shortcutsData-as-a-ProductDomain-based data meshFabric-as-a-ServiceAI-ready data

Status quo · late 2026

What has changed — and why it matters now

The lake alone is an anti-pattern

Raw object storage without table formats is now considered a mistake for production. Lakehouse — ACID tables on open formats — is the default.

Open formats won

Iceberg and Delta interoperate; Fabric, Databricks, Snowflake and AWS all read them. Storage is portable; compute is a choice, not a lock-in.

Lakehouse feeds AI

Vector indexes, feature tables and unstructured documents live beside BI tables under one catalogue — the lakehouse is where AI-ready data lives.

How we approach it

Our method for Data Lake Architecture

01AssessCurrent estate, workloads, data domains, cost, sovereignty and skills.

02DesignPlatform choice, table format, medallion layers, catalogue, security model, FinOps guardrails.

03BuildLanding zones, ingestion frameworks, transformation standards, CI/CD, IaC.

04ProductiseDomain data products with contracts, SLAs and ownership.

05RunFabric-as-a-Service: operations, optimisation, cost management, evolution.

In depth

Architecture decisions

Table format

Iceberg vs Delta vs Hudi — engine support, time travel, schema evolution, catalogue.

Medallion layers

Bronze (raw), Silver (conformed), Gold (semantic / products) with clear contracts.

Catalogue & security

Unity Catalog, Purview, Glue, Dataplex — lineage, fine-grained access, tags.

Compute strategy

Serverless SQL, Spark, streaming engines; right-sizing and autoscaling.

Zero-copy sharing

OneLake shortcuts, Delta Sharing, Iceberg REST catalogues across clouds and partners.

What you get

Deliverables

Fabric-as-a-ServiceETL / ELT pipelinesLakehouse designDomain-based data meshData-as-a-Product models

Technology-agnostic

Platforms we work across

We design for the outcome first and choose the platform second — across Microsoft, AWS and Google Cloud, and every major AI provider.

Microsoft Fabric & OneLakeDatabricksSnowflakeAWS S3 / Glue / Athena / RedshiftGoogle BigQuery / BigLake / DataplexAzure Data Lake Storagedbt

More Data services

Data Engineering Data Analytics Data Science Data Governance Data Mesh, Fabric & API Access

Ready to talk about Data Lake Architecture?

One email reaches the architects and engineers who will do the work. We reply within one working day.

info@uspc.co.uk Contact page

Data Lake Architecture— lakehouse, open formats