1.1.2. The Build vs. Buy Spectrum
💡 First Principle: Every data engineering decision falls somewhere on a spectrum from "build custom" to "use managed services." Building gives you control but costs time and expertise. Managed services give you speed but constrain your options. There is no universally correct position—only trade-offs appropriate to your context.
| Approach | Control | Speed | Expertise Required | Cost Model |
|---|---|---|---|---|
| Custom Code (Build) | Maximum | Slow | High | Dev + Ops |
| Open Source Tools | High | Medium | Medium | Ops + Support |
| Managed Services (Buy) | Limited | Fast | Low | Consumption |
Microsoft Fabric sits firmly in the "managed services" category. The exam tests your ability to work within Fabric's constraints, not around them.
More precisely, Fabric is SaaS: one tenant-wide service with a single identity (Microsoft Entra ID), one lake (OneLake) and one capacity model, and Microsoft operates the underlying infrastructure—including availability-zone redundancy in supported regions, with cross-region disaster recovery as an opt-in capacity setting. Azure Synapse Analytics is PaaS: you provision and manage its workspaces, pools and storage accounts as separate Azure resources.
⚠️ Exam Trap: The exam rarely presents "write custom code" as the correct answer when a Fabric-native solution exists. If Fabric can do something out-of-the-box, that's almost always the right answer—even if custom code would be more flexible.
