← AZ-305: Azure architecture and production decisions
04 / 8 · 25 MIN

Storage and data integration

Relate distribution, access, retention, and integration paths.

Concept and mechanism

The partition key influences where Cosmos DB load concentrates. Spare aggregate capacity does not prevent a heavily used key from limiting the workload. In Blob Storage, access tier should match required read timing: Archive requires rehydration before online reading. Redundancy is another dimension. ZRS distributes copies across zones within a region; geographic copying involves another scope and must respect supplied location and recovery requirements. No choice should be made solely on storage unit price.

Guided application

For hybrid integration, confirm the path to the private source and where the runtime executes. A self-hosted integration runtime can provide that access with its own availability and credentials to manage. For immutable retention, confirm scope and duration before locking. A locked time-based policy does not permit shortening the period; the architect implements approved requirements without inventing legal obligations.

IN PRACTICE

An inexpensive recovery file in Archive can fail the immediate-read requirement during an incident.

Common pitfalls

Confusing tier with redundancy; ignoring hot partitions; locking retention before agreement across data, business, and cost owners.

Related topics: Continuity and demonstrated recovery · Compute and operational responsibility

Take this idea with you

Storing also means deciding how to access, integrate, recover, and retire data.

Create account

Reference: Cosmos DB partitioning · AZ-305 objectives 2026-04-17