CSA-in-a-Box¶

Azure-native reference implementation of Microsoft's "Unify your data platform" Cloud Adoption Framework guidance.

CSA-in-a-Box assembles Azure PaaS services and open-source tooling into an opinionated, end-to-end data platform that delivers Data Mesh, Data Fabric, and Data Lakehouse capabilities today — designed for environments where Microsoft Fabric is not yet GA (Azure Government), and as an incremental on-ramp for teams building toward Fabric adoption.

Fork it, deploy it, customize it. Production-grade Bicep + reference code you own and operate.

Start the 30-min tour Quickstart (5 min) View on GitHub

Start here¶

Four pages cover the full path from "what is this?" to "deployed in production."

Why teams use it¶

The principles behind the platform¶

CSA-in-a-Box is not just a collection of Bicep modules — it is a working implementation of three converging data architecture paradigms that together define what "cloud-scale analytics" means in practice.

Data Mesh — domain-oriented ownership¶

Data Mesh treats data as a product owned by the domain that produces it, not a centralized team. In CSA-in-a-Box:

• Domain-oriented Data Landing Zones — each business domain (finance, sales, inventory) owns its own DLZ subscription with its own ADLS Gen2 storage, compute, and pipelines.
• Self-serve data infrastructure — domain teams deploy from shared Bicep modules and dbt project templates without waiting on a central platform team.
• Federated computational governance — Purview enforces classification, lineage, and access policies across all domains from the central DMLZ, while domain teams retain ownership of their data products.
• Data product contracts — YAML-defined contracts specify schema, SLAs, freshness guarantees, and ownership, enabling consumers to discover and trust data across domain boundaries.

Data Fabric — unified metadata & governance¶

Data Fabric provides an integrated layer of metadata, governance, and automation that spans all data assets regardless of where they live. In CSA-in-a-Box:

• Microsoft Purview is the unified catalog — scanning, classifying, and tracking lineage across ADLS Gen2, Databricks, Synapse, Azure SQL, and Cosmos DB.
• Automated governance — sensitivity labels, access policies, and compliance controls propagate across data assets without manual tagging.
• Cross-domain data discovery — the Data Marketplace API enables self-service search, access requests, and data product registration.
• Lineage from source to dashboard — end-to-end tracking from ingestion (ADF) through transformation (dbt/Spark) to consumption (Power BI, APIs).

Data Lakehouse — Delta Lake medallion¶

The Data Lakehouse unifies data lakes (scalable, open storage) with data warehouses (ACID transactions, schema enforcement, BI performance). In CSA-in-a-Box:

• Delta Lake on ADLS Gen2 — open-format, ACID-compliant tables on low-cost cloud storage, readable by Spark, Synapse Serverless SQL, and Power BI without data movement.
• Medallion architecture (Bronze / Silver / Gold) — raw → validated → business-ready, with quality gates enforced by Great Expectations at each transition.
• Unified batch and streaming — the same Delta tables serve both batch pipelines (ADF + dbt) and streaming workloads (Event Hubs + Spark Structured Streaming).
• Compute diversity — Databricks, Synapse Spark, and Synapse Serverless SQL all query the same lakehouse, so teams pick the engine that fits their workload without duplicating data.

Why "one-stop shop"? Most reference implementations cover one of these paradigms. CSA-in-a-Box implements all three — plus AI/ML integration, compliance mappings for six regulatory frameworks, 11 migration playbooks, 18 end-to-end vertical examples, and production runbooks — in a single, fork-ready repository.

Choose your path¶

Use Fabric, or use this?¶

CSA-in-a-Box is not a blanket substitute for Microsoft Fabric. For most Azure Commercial greenfield workloads where Fabric is GA in the region, Fabric is the right answer.

For the full decision logic — including Fabric vs. Databricks vs. Synapse — see the Fabric vs. Databricks vs. Synapse decision tree and ADR-0010: Fabric Strategic Target.

For the full capability matrix and Fabric-equivalent mapping, see Architecture → What's Included.

For deep-dive Fabric content — tutorials, feature guides, best practices, POC agendas, and ready-to-run notebooks — see the Supercharge Microsoft Fabric companion site.

See also:

• → Next: Getting Started (30-min tour)
• ⌂ Index: Documentation home