Backup and Disaster Recovery Design
AZ-305 AZ-305 · Design business continuity solutions
Business continuity is the discipline of surviving the worst day: a datacentre fire, a ransomware attack, a regional outage, or an administrator's mistake. AZ-305 tests it through the twin lenses of recovery objectives — how much data can you lose (RPO) and how long can you be down (RTO) — and the Azure services that deliver them: Azure Backup and Azure Site Recovery.
RPO and RTO: the numbers that drive every decision
Recovery point objective (RPO) is the maximum acceptable data loss, measured in time: an RPO of one hour means the business can lose at most one hour of data. Recovery time objective (RTO) is the maximum acceptable downtime: an RTO of four hours means the service must be back within four hours. Every backup and DR design starts here, because the objectives determine the technology. A 15-minute RPO needs continuous or very frequent replication; a 24-hour RPO can be met with daily backups. A near-zero RTO needs hot standby or active-active design; a 24-hour RTO can be met with restore-from-backup. The exam tests this relentlessly: read the RPO/RTO in the scenario first, then choose the technology that meets them at the lowest cost. Over-engineering (continuous replication for a 24-hour RPO) is as wrong as under-engineering.
Azure Backup: what it protects and how
Azure Backup provides backup for Azure VMs, on-premises servers (via the MARS agent or Azure Backup Server), SQL Server in Azure VMs, Azure SQL Database (via long-term retention), Azure Files, and Blob Storage (via operational backup and vaulted backup). Backups are stored in a Recovery Services vault (or a Backup vault for newer workloads), with storage redundancy options: LRS for the vault in the primary region, GRS for a copy in the paired region, and zone-redundant storage (ZRS) for vaults in zone-enabled regions. Key design points: back up to a vault in the same region as the workload for fast restores, with GRS for regional protection; use backup policies to define frequency and retention (daily/weekly/monthly/yearly tiers); and for ransomware protection, enable soft delete on the vault, use immutable vaults (which prevent backup deletion even by a compromised admin), and require multi-user authorisation (MUA) via a resource guard for destructive operations. Instant restore keeps VM snapshots for fast recovery of recent states without waiting for vault transfer.
Azure Site Recovery: disaster recovery by replication
Azure Site Recovery (ASR) replicates workloads to a secondary region (or to Azure from on-premises) so they can be failed over when the primary region is unavailable. It supports Azure VMs, VMware and Hyper-V VMs, and physical servers. Design points: choose the target region (usually the paired region for proximity and platform guarantees); plan the recovery plan — the orchestrated sequence of failover steps, including boot order, IP addressing, and runbook automation; test with test failovers (which validate the plan in an isolated network without affecting production); and set the RPO threshold alert so you know when replication lag exceeds the objective. ASR replication for Azure VMs uses a crash-consistent snapshot frequency of as low as 30 seconds, with app-consistent snapshots configurable. Remember the distinction the exam probes: Backup restores data to a point in time; Site Recovery replicates running workloads for failover. They complement each other — backup protects against data loss and corruption, Site Recovery protects against site loss.
Backup for databases and unstructured data
Databases need their own strategy. Azure SQL Database has built-in automated backups with point-in-time restore and long-term retention (covered in the relational guide). SQL Server on VMs uses Azure Backup with transaction-log backups for point-in-time restore. For Blob Storage, use operational backup (leveraging soft delete and versioning, with policy-based retention) or vaulted backup (copies data to a Backup vault for isolated, long-term protection). For Azure Files, Azure Backup provides snapshot-based backup with file-level restore. The exam often asks which backup type fits a scenario: vaulted backup when isolation from the source account matters (ransomware protection), operational backup when fast, low-cost recovery is the priority.
Designing the DR plan
A complete DR design includes: the RPO/RTO per workload tier (not every workload needs the same objectives — tier them by criticality); the replication or backup mechanism per tier; the network design for failover (VNets in the target region, DNS updates via Traffic Manager or Front Door); the recovery plan with runbooks; and the testing schedule. Test failovers should be run regularly — an untested DR plan is a hope, not a plan. Cost control comes from tiering: hot standby only for the workloads whose RTO demands it, with pilot-light or backup-and-restore for the rest.
Testing, documentation, and cost optimisation
A disaster recovery plan that has never been tested is a hope, not a plan. Schedule test failovers at least quarterly for critical workloads, and run full game days annually where the team executes the recovery plan end to end, including communication, DNS changes, and stakeholder notification. Document the plan with clear ownership: who declares a disaster, who executes the failover, who communicates with customers, and what the decision criteria are. Keep the documentation outside the affected region — a recovery plan stored only in the failed region is useless. For cost optimisation, tier ruthlessly: hot standby only for workloads whose RTO demands it, pilot light (minimal infrastructure running, scaled up on failover) for the middle tier, and backup-and-restore for the rest. Review recovery objectives annually, because business requirements change and an RTO that was essential three years ago may now be over-engineered. Finally, remember that backups and replication protect against infrastructure failure and data loss — they do not protect against application bugs that corrupt data and replicate the corruption. Combine infrastructure recovery with application-level safeguards such as soft delete, versioning, and point-in-time restore.
Protecting Microsoft 365 and SaaS data
Azure Backup covers Azure workloads, but many organisations forget their SaaS data. Microsoft operates on a shared-responsibility model: it protects the service, but the customer is responsible for their data in it. Microsoft 365 Backup (and third-party backup solutions) protect Exchange mailboxes, SharePoint sites, OneDrive accounts, and Teams data against accidental deletion, malicious destruction, and ransomware — with fast restore times measured in hours, not days. The exam occasionally tests this boundary: a scenario asking about backup for SharePoint or Exchange is not an Azure Backup question, because Azure Backup does not cover Microsoft 365 workloads. Include SaaS data in the business continuity plan, with its own RPO and RTO, and verify restore procedures just as you would for Azure workloads.
Exam traps to watch for
- Choosing Site Recovery when the scenario only needs point-in-time data restore — backup is the cheaper answer for data loss; Site Recovery is for site failover.
- Ignoring the RPO/RTO in the scenario — the objectives determine the technology; over-engineering is as wrong as under-engineering.
- Forgetting ransomware protections on backup vaults — soft delete, immutable vaults, and multi-user authorisation are the exam's expected answers.
- Assuming backup storage redundancy replaces geo-replication — GRS backup storage protects the backups; it does not fail over the workload.
Practice questions
Test yourself on this guide. Answers are checked instantly, with a full explanation for each question.
Q1.A company's critical application has an RPO of 30 minutes and an RTO of 2 hours for a regional outage. The application runs on Azure VMs with Azure SQL Database. What should the architect recommend?
Explanation: Site Recovery replicates the VMs continuously for failover within the 2-hour RTO, and a SQL failover group provides automatic database failover — together they meet both objectives for a regional outage. Option A is wrong because daily backups cannot meet a 30-minute RPO. Option C is wrong because geo-redundant disk storage does not fail over running VMs. Option B is wrong because zone redundancy protects against datacentre failure within a region, not a regional outage.
Exam objective: AZ-305 Domain 3: Design solutions for backup and disaster recovery
Q2.A company was hit by ransomware that encrypted production data and also deleted the backups in the Recovery Services vault. What should the architect recommend to prevent backup deletion in future attacks?
Explanation: Soft delete retains deleted backups for recovery, immutable vaults prevent backup deletion entirely, and multi-user authorisation requires a second approver for destructive operations — together they stop an attacker (even with admin credentials) from destroying backups. Option D is wrong because more frequent backups are still deletable. Option C is wrong because a storage account in another region is equally deletable if the attacker has credentials. Option B is wrong because encryption protects data confidentiality, not deletion.
Exam objective: AZ-305 Domain 3: Design solutions for backup and disaster recovery
Q3.A company needs to restore individual files from a file share after users accidentally delete them, and needs backups retained for 90 days. Which solution should the architect recommend?
Explanation: Azure Backup for Azure Files provides snapshot-based backup with file-level restore and configurable retention — exactly the requirement. Option D is wrong because Site Recovery replicates VMs, not file shares, and is for disaster recovery, not file restore. Option C is wrong because blob versioning applies to blobs, not Azure Files shares. Option B is wrong because geo-redundancy protects against regional failure, not accidental deletion.
Exam objective: AZ-305 Domain 3: Design solutions for backup and disaster recovery
Q4.An architect is designing backup for Azure VMs. The business requires a 24-hour RPO and a 12-hour RTO, and wants to minimise cost. What should the architect recommend?
Explanation: A daily backup policy meets a 24-hour RPO, restore from backup meets a 12-hour RTO, and locally redundant vault storage is the cheapest option — the objectives do not justify replication or geo-redundancy. Option B is wrong because continuous replication is over-engineered and costly for these objectives. Option A is wrong because hourly snapshots and GRS exceed what the objectives require. Option D is wrong because hot standby is the most expensive option and unnecessary for a 12-hour RTO.
Exam objective: AZ-305 Domain 3: Design solutions for backup and disaster recovery
Q5.A company wants to validate its disaster recovery plan without disrupting production. What should the architect recommend?
Explanation: A test failover creates the recovered environment in an isolated network, validating the recovery plan without touching production. Option B is wrong because a full failover during business hours would disrupt production. Option C is wrong because restoring to production risks overwriting live data. Option A is wrong because toggling replication validates nothing about the recovery plan.
Exam objective: AZ-305 Domain 3: Design solutions for backup and disaster recovery
Q6.A SQL Server database running on an Azure VM was corrupted at 14:00. The last full backup was at 02:00 and transaction log backups run every 15 minutes. The database must be restored to 13:45. What should the architect recommend?
Explanation: Point-in-time restore replays transaction log backups on top of the full backup to reach any point within the log chain — 13:45 is achievable with 15-minute log backups. Option B is wrong because it loses 12 hours of data unnecessarily. Option D is wrong because Site Recovery fails over the VM to a replicated copy, which would also contain the corruption. Option C is wrong because geo-redundant storage is a copy of the same backups, not a different restore point.
Exam objective: AZ-305 Domain 3: Design solutions for backup and disaster recovery
Q7.A company stores critical blobs in a storage account and wants backup copies isolated from the source account so that a compromise of the account cannot destroy the backups. Which backup approach should the architect recommend?
Explanation: Vaulted backup copies blob data into a separate Backup vault, isolated from the source storage account — a compromise of the account cannot reach the backups. Option C is wrong because operational backup relies on features within the same account, so an attacker with account access could delete the protected data. Option D is wrong because GRS replicates the account's data, including deletions, to another region. Option B is wrong because tiering changes storage cost, not backup isolation.
Exam objective: AZ-305 Domain 3: Design solutions for backup and disaster recovery
Q8.A multi-tier application must fail over to a secondary region in a specific order: database first, then application servers, then web servers. How should the architect implement this?
Explanation: A recovery plan orchestrates failover in defined groups with sequencing, scripts, and manual steps — it enforces the database-first order. Option A is wrong because separate vaults do not sequence failover. Option C is wrong because Traffic Manager routes DNS traffic; it does not orchestrate VM failover order. Option B is wrong because auto-failover groups apply to Azure SQL Database, not to VM tiers.
Exam objective: AZ-305 Domain 3: Design solutions for backup and disaster recovery
Q9.A company replicates VMs to a paired region with Azure Site Recovery. The operations team wants to be alerted if replication lag exceeds the 30-minute RPO. What should the architect recommend?
Explanation: Site Recovery lets you configure an RPO threshold alert that fires when replication lag exceeds the specified threshold — the purpose-built mechanism. Option A is wrong because VM CPU has no relationship to replication lag. Option C is wrong because the activity log records control-plane operations, not replication health. Option D is wrong because resource health reports platform issues, not replication lag against an RPO.
Exam objective: AZ-305 Domain 3: Design solutions for backup and disaster recovery