Best Practices¶
Nine field-tested guides for running cloud-scale analytics + AI on Azure. Each one is independent — read the ones relevant to your role.
| Guide | When to read | Length |
|---|---|---|
| Medallion Architecture | Designing your bronze/silver/gold lakehouse | 664 lines |
| Data Engineering | Authoring ADF + dbt + Spark pipelines | 800 lines |
| Data Governance | Setting up Purview, contracts, lineage, classification | 573 lines |
| Security & Compliance | Hardening identities, secrets, network, encryption | 654 lines |
| Infrastructure as Code & CI/CD | Bicep, what-if, GitHub Actions, environment promotion | 657 lines |
| Cost Optimization | Tagging, reserved capacity, auto-pause, FinOps | 518 lines |
| Monitoring & Observability | Log Analytics, Workbooks, OTel, SLI/SLO | 520 lines |
| Performance Tuning | Spark configs, Synapse SQL pools, AI Search shards | 705 lines |
| Disaster Recovery | RPO/RTO targets, geo-replication, runbook drills | 521 lines |
How to use these¶
Each guide follows the same structure:
1. The problem (1-2 paragraphs)
2. The opinionated answer (this is what we do)
3. The reasoning (why — usually links to an ADR)
4. The mechanics (commands, code, configs)
5. The trade-offs (what we gave up to make this choice)
6. The escape hatches (when this advice does NOT apply)
If a guide ever reads as "do X because everyone does X," that's a bug — open an issue.
Related¶
- ADRs — the 22 specific decisions these best-practices are built on
- Reference Architectures — how the pieces fit together
- Patterns — implementation patterns for specific scenarios
- Decision Trees — quick "which option do I pick" guides