ISC · Information Systems and Data Management · 6 practice questions
Multi-AZ vs. Multi-Region: Regional Outage Protection
Multi-AZ protects availability within one region, not against loss of that entire region. Below, compare one retailer’s deployment with alternatives that add either cross-region backups or a cross-region database replica.
Try one first
Hint
Separate protection against local component failures from protection against a full regional outage, and do not treat backups as the same as live failover capability.
Answer C. The architecture provides resilience against localized failures within the region (instance or single-AZ failures) because of redundant application servers and synchronous multi-AZ database failover. However, a regionwide outage disables all availability zones in that region; without a live cross-region replica, preprovisioned compute in the second region, or automated DNS failover, the company would face substantial downtime and potential loss of transactions since the last backup (an RPO gap). Backups in another region help recovery but do not deliver near-immediate continuity.
Why not A: Attractive because synchronous multi-AZ sounds robust, but it protects against AZ-level failures within the same region. If the entire region is unavailable, all zones are affected and regional replication or DR is still required for continuity.
Why not B: Tempting because candidates conflate offsite backups with high-availability failover. Backups support restoration after an outage but do not provide automatic live failover or continuity; without replication, preprovisioned resources, and automated cutover, interruption and data loss since the last backup remain likely.
Why not D: A reasonable-sounding minimization, but it understates the impact of a full regional outage. The described controls mitigate local availability issues, not the loss of an entire region; the central residual risks are downtime and recovery timing, not merely degraded performance.
Side by side
| Multi-AZ only | Multi-AZ + cross-region backups | Multi-AZ + cross-region database replica | |
|---|---|---|---|
| Retailer’s starting deployment | Web/app servers across two AZs; Multi-AZ database; nightly backups in Region 1. | Same starting deployment. | Same starting deployment. |
| One change from the starting deployment | None. | Copy nightly backups to Region 2. | Add a live database replica in Region 2; no application compute is preprovisioned. |
| Single-zone outage | The other zone supports continued service. | The other zone supports continued service. | The other zone supports continued service. |
| Complete Region 1 outage | Both local zones become unavailable. | Both local zones become unavailable. | Both local zones become unavailable. |
| Database recovery source outside Region 1 | None in this design. | Copied nightly backups in Region 2. | Live database replica in Region 2. |
| Application compute in Region 2 | Not preprovisioned. | Provision compute before service resumes. | Provision compute before service resumes. |
| Recovery after Region 1 fails | No recovery path outside Region 1 is established. | Provision application and database compute in Region 2, restore a backup, and cut over traffic. | Provision application compute in Region 2, promote the replica, and cut over traffic. |
Key points
- Remember: Two availability zones in one region still share exposure to a complete regional outage.
- Cross-region backups provide a recovery source, not a running application.
- A cross-region database replica does not supply the application compute needed to resume service.
How the exam traps you
- Treating synchronous Multi-AZ database failover as protection against losing the entire region. Check where the standby resides. Another zone in Region 1 is still inside the failed region.
- Calling cross-region backups a complete high-availability deployment. Backups must be restored into usable infrastructure before the application can resume.
- Assuming a cross-region database replica means the application is already running there. The replica supplies database infrastructure, but this design still requires application compute in Region 2.
Question 2
Hint
For each requirement, identify the supporting technology: RTO, how traffic is distributed and how quickly failover occurs; RPO, replication mode or backup frequency; elasticity, whether scaling is automatic and cross-site.
Answer C. An active-active multi-region deployment serves traffic from multiple sites, enabling near-immediate failover and very low RTO; synchronous cross-region replication preserves committed transactions across regions, giving very low RPO; and cloud-based auto-scaling provides rapid, automated capacity increases during traffic spikes. Together these directly satisfy all three requirements.
Why not A: This choice is tempting since a hot-standby and secondary site appear redundant. It fails because asynchronous replication can lose recent transactions on a site outage (worse RPO), DNS-based failover is typically slower and subject to caching delays (longer effective RTO), and it does not provide automated cross-site auto-scaling for rapid elasticity.
Why not B: This option tempts because it preserves existing infrastructure and uses backups and a local load balancer for some availability. It fails because nightly backups permit up to 24 hours of data loss (poor RPO), manual failover causes prolonged downtime (poor RTO), and there is no automated scaling for rapid traffic spikes.
Why not D: Adding servers and doing backups may improve capacity and data retention superficially, which makes this option tempting. It fails because a single data center remains a single point of failure (poor RTO), weekly tape backups allow substantial data loss (very poor RPO), and there is no automated cross-site scaling to handle sudden spikes.
Question 3
Hint
For each option, ask: does it eliminate the shared-storage single point of failure, and does it guarantee both the 15-minute RTO and zero loss of committed transactions (RPO = 0)?
Answer B. This single initiative is the only approach that addresses both constraints together. Independent compute and storage at the secondary site removes the shared-storage single point of failure (protecting against storage-array or full-site loss). Synchronous replication preserves committed transactions at the secondary (RPO = 0), and tested automated failover provides a repeatable path to meet the 15-minute RTO. The other choices either retain a replication lag, do not create an independently usable secondary copy for a site outage, or rely on backups/restores that cannot guarantee transactional zero-loss and rapid automated recovery.
Why not A: Tempting because adding compute and automating cutover addresses recovery speed (RTO) and makes the secondary usable more quickly. However, keeping asynchronous replication with a 5-minute lag still risks losing recently committed orders in a primary-site outage, so it fails the no-data-loss (RPO = 0) requirement.
Why not C: Tempting because it reduces the likelihood of a local storage hardware failure and improves primary-site availability at lower cost. However, it does not create an independently usable copy at the secondary site and therefore cannot guarantee recovery after a full primary-site outage or eliminate the risk of losing committed transactions if the site becomes unavailable.
Why not D: Tempting because more frequent backups and scripted restores can reduce potential data loss and speed recovery relative to ad hoc processes. However, backups/snapshots are not a synchronous, application-consistent replication mechanism; restore times and transactional consistency cannot be guaranteed to meet a 15-minute RTO or provide RPO = 0 for committed orders.
Question 4
Hint
Separate the two targets before judging the design: one measures allowable data loss, and the other measures how long restoration can take.
Answer C. The binding objective is the 2-hour RTO. Rebuilding the web/app tiers (~30 minutes) plus manual DNS (~10 minutes) fits within that window, and copying logs every 5 minutes meets the 15-minute RPO. However, a tested 5-hour database restore in Region 2 exceeds the 2-hour RTO, so the primary infrastructure problem is the slow database recovery.
Why not A: Tempting because manual steps add operational risk and delay. However, the DNS cutover is only about 10 minutes, small relative to the 2-hour RTO, so it is not the primary constraint when a database restore takes 5 hours.
Why not B: Tempting because candidates often conflate RPO with the need for continuous replication. But an RPO of 15 minutes allows up to 15 minutes of potential data loss; copying logs every 5 minutes implies at most a ~5-minute loss and therefore meets the stated RPO. The scenario's decisive failure is the long restore time, an RTO issue.
Why not D: Tempting because active-active is often recommended for high availability. But it is an overstatement: a standby or restore-based design can be acceptable if it meets the RTO/RPO. The problem in this fact pattern is that the specific restore-based approach fails the RTO due to the slow database recovery, not that active-active is the only valid architecture.
Question 5
Hint
Separate the scope of the failure the system must survive from the time and data objectives that apply after failover.
Answer D. The decisive issue is the failure domain the system must survive. A multi-availability-zone design handles zone-level failures but cannot guarantee continued operation through a full-region outage; only a multi-region deployment can meet that scope requirement. Because the stem says either design can meet the 30-minute downtime and 10-minute data-loss targets for the failure scenarios they cover, RTO and RPO are not the governing factor here.
Why not A: Backup retention matters for how far back you can restore data, but it does not determine whether the live production environment can continue operating through a regional cloud failure. This item is about required fault-domain coverage, not data-retention windows.
Why not B: This distractor plays on RPO concerns and replication; it's tempting because data-loss limits often drive DR design. However, the stem already states both architectures can be engineered to meet a 10-minute data-loss target for the scenarios they cover, so the unresolved question is the outage scope (region vs. zone).
Why not C: Downtime tolerance (RTO) is another plausible driver of architecture complexity, so this option is attractive. But like the RPO option, the stem explicitly says either design can meet the 30-minute downtime target for its covered failure scenario; only the required outage scope distinguishes the needed architecture.
Question 6
Hint
Which choice can provide near-instant recovery with near-real-time replication, elastic bursting, in-country data residency, and less ongoing operational work for a small IT team?
Answer C. Option C best satisfies all constraints: an in-country active-active cloud deployment with managed services (regional traffic management, auto-scaling stateless tiers, and strongly consistent or near-synchronous database replication) can provide elastic scaling and the sub-5-minute RTO and near-1-second RPO required, while enabling in-country data residency and shifting routine operational tasks to the provider. Equivalent on-premises solutions that meet these tight RTO/RPO targets would require substantially more capital and ongoing staffing, conflicting with the stem's limited IT resources.
Why not A: This is tempting as a lower-cost DR approach that preserves an offsite standby. It fails the stem's recovery objectives because asynchronous replication typically cannot guarantee an RPO of ≤1 second, and active-passive failover procedures commonly involve manual steps or longer cutover times that exceed a 5-minute RTO.
Why not B: This option is tempting because synchronous active-active on-premises architecture can deliver very low RPO and fast failover while keeping full physical control of data. It fails here because standing up and operating a second full datacenter with synchronous replication and active-active clustering is capital- and staff-intensive and adds complex scaling and operations responsibilities, contradicting the requirement to minimize ongoing operational burden given limited IT staff.
Why not D: This option is tempting for low upfront cost and familiar controls. It fails because local redundancy and nightly backups are recovery measures, not instant failover; they cannot provide near-zero data loss or sub-5-minute recovery and therefore cannot meet the stated RPO and RTO.
Common questions
Does Multi-AZ protect against a regional outage?
No. In this comparison, both availability zones are in Region 1, so a complete Region 1 outage makes both unavailable.
Are cross-region backups the same as multi-region high availability?
No. The retailer has a recovery source outside Region 1, but it must provision infrastructure and restore the backup before resuming service.
Does a cross-region database replica make the whole application region-resilient?
Not by itself. The retailer must also provision application compute in Region 2, promote the replica, and redirect traffic.
Practice ISC like the real exam
The free ChatCPA simulator: real exam layout, timed testlets, starting with a question on this topic. No account needed to start.
Open the free simulator →More on IT infrastructure
Questions from the ChatCPA bank of 17,658 CPA exam questions, each with a written reason for every wrong answer. ChatCPA is built by Nicholas Miller, CPA (Oregon #14907). How these pages are made. Spot an error? Tell us.