Design Studio — Design for high availability

This studio turns the topic into an architecture exercise. Produce a recommendation that another engineer could challenge, implement, and validate. There is no credit for a list of Azure products without flows, constraints, failure behaviour, and trade-offs.

Client brief

Fabrikam is modernising a production workload in this domain: Remove failure domains and add detection, traffic steering, state continuity, capacity, and recovery automation. The workload serves internal teams and external customers, carries regulated data, and must remain supportable by a small platform team. The client has supplied incomplete requirements; part of the exercise is to state assumptions and identify questions that would materially change the recommendation.

The initial requirements are:

1. The design must address datacenter failure with explicit assumptions and measurable acceptance evidence.
2. The design must address regional failure with explicit assumptions and measurable acceptance evidence.
3. The design must address global http traffic with explicit assumptions and measurable acceptance evidence.
4. The design must address stateful continuity with explicit assumptions and measurable acceptance evidence.
5. The solution must use Microsoft Entra identities, private connectivity where justified, infrastructure as code, and observable health signals.
6. The design must state a recovery path, operational owner, cost driver, and one deliberate compromise.

Candidate decision anchors

These are starting points, not an answer key. You may select a different service when a stated assumption or constraint justifies it. Every deviation must identify the requirement it serves and the new operational cost it introduces.

Required submission

1. A one-page architecture diagram with a text equivalent.
2. A decision record containing context, decision, alternatives, consequences, and validation evidence.
3. A requirement-to-control matrix that maps every hard constraint to a component or operating process.
4. A failure table covering component, zone, region, identity, network, and operator-error failures where applicable.
5. Three validation tests: one functional, one reliability/security, and one operational or cost test.

Review traps

• Multiple instances in one zone are not zonally resilient.
• An SLA is not an architecture calculation.
• Redundancy without health probes is not self-healing.

Assessment rubric

Reflection

• Which requirement eliminated the most attractive alternative?
• Which assumption creates the greatest residual risk?
• What would you test in a pilot before approving production deployment?
• How would the recommendation change if the operations team doubled in size or the recovery objective tightened?

Source and freshness

Grounded in the current AZ-305 skills outline and both attached course sources. Current service contracts must be verified in Microsoft Learn during the studio. Reviewed 2026-08-02.