ISC · Information Systems and Data Management · 10 practice questions
RTO vs RPO: Synchronous vs Asynchronous Replication
Recovery time must meet RTO, and maximum potential transaction loss must meet RPO. Below, compare a base scenario with versions that change only replication or failover.
Try one first
Hint
Consider whether a transaction can be acknowledged before the remote site confirms it and whether failover remains manual or becomes automatic.
Answer B. Synchronous replication requires a remote acknowledgement before a commit is confirmed, so the chance of data loss is effectively eliminated (RPO near zero). Automatic failover in an active‑active setup removes manual switchover delays, reducing recovery time (shorter RTO). The necessary cross‑site acknowledgement round trip increases transaction commit latency.
Why not A: This is tempting because active‑active can balance load and lower queuing delays, suggesting lower latency. It fails because synchronous commits require remote acknowledgement (which adds commit latency) and automatic failover usually reduces manual switchover time (so RTO is unlikely to remain unchanged).
Why not C: This appeals to candidates who expect some residual replication lag after a change. It is incorrect because properly configured synchronous replication forces commits to be confirmed on both sites, effectively reducing replication lag to near zero rather than leaving RPO meaningfully above zero.
Why not D: This distractor plays on the idea that tighter coupling increases complexity and recovery time. While the latency claim is plausible, the RTO claim is wrong for a correctly implemented active‑active design with automatic failover, which generally shortens recovery time rather than increasing it.
Same scenario, one fact changes
Base case
NovaPay runs a primary database at Site A and a standby at Site B. Its targets are RTO ≤ 15 minutes and RPO ≤ 1 minute. Asynchronous replication has a maximum lag of 5 minutes. After Site A fails, automatic failover restores full service in 10 minutes. Site B is the only recovery source.
Answer: RTO is met; RPO is not.
Recovery takes 10 minutes, within the 15-minute RTO. Potential transaction loss reaches 5 minutes, exceeding the 1-minute RPO.
Before you open each one, predict the answer.
Change 1Replace asynchronous replication with synchronous replication that durably stores committed transactions at both sites.
Answer: Both RTO and RPO are met.
Recovery remains 10 minutes. Durable commits at both sites support zero committed-transaction loss, within the 1-minute RPO.
Change 2Replace 10-minute automatic failover with manual failover that restores full service in 25 minutes.
Answer: Neither RTO nor RPO is met.
Recovery takes 25 minutes, exceeding the 15-minute RTO. The unchanged 5-minute maximum lag still exceeds the 1-minute RPO.
Change 3Reduce maximum asynchronous replication lag from 5 minutes to 1 minute.
Answer: Both RTO and RPO are met.
Recovery remains 10 minutes. Maximum potential transaction loss is 1 minute, exactly the permitted RPO.
Key points
- Remember: Fast recovery does not prove that recent transactions are safe; test RTO and RPO separately.
- Compare maximum replication lag, not average lag, with the permitted transaction-loss window.
- A recovery time or loss window exactly equal to its stated maximum meets that target.
How the exam traps you
- Treating automatic failover as proof that no transactions can be lost. Check replication separately. Automatic failover can restore service while recent committed transactions are missing.
- Concluding that meeting either RTO or RPO means the design meets both. Make two comparisons. Both must pass.
- Assuming asynchronous replication always fails RPO. Compare maximum lag with the allowed loss window. A one-minute lag meets a one-minute RPO.
Now the same facts as questions
Each question changes one fact from the one before. Watch which change flips the answer.
Question 1
Answer B. Ten minutes meets RTO; five minutes of potential loss exceeds RPO.
Why not A: Automatic failover does not eliminate data loss from asynchronous replication.
Why not C: Compare lag with RPO, not with recovery time.
Why not D: Do not add lag to recovery time. Each has a separate target.
Question 2
Answer C. Ten-minute recovery meets RTO; durable commits at both sites meet RPO.
Why not A: The five-minute asynchronous loss window no longer applies.
Why not B: Recovery remains 10 minutes, within the RTO.
Why not D: Nothing in the changed fact prevents automatic failover.
Question 3
Answer D. Both comparisons fail: 25 exceeds 15, and 5 exceeds 1.
Why not A: Having a standby does not establish compliance with either time limit.
Why not B: Twenty-five minutes exceeds the 15-minute RTO.
Why not C: Replication remains asynchronous with a five-minute maximum lag, exceeding RPO.
Question 4
Answer A. Recovery is within 15 minutes, and maximum lag equals the permitted one-minute loss window.
Why not B: Asynchronous replication can meet a nonzero RPO when maximum lag stays within its limit.
Why not C: The one-minute limit is RPO. Recovery is compared with the 15-minute RTO.
Why not D: The targets use ≤, so equality meets the limit.
Question 2
Hint
Separate the recovery-time requirement from the allowable-data-loss requirement, and test each one against the facts independently.
Answer A. The commitment defines two separate thresholds: an RTO of 60 minutes and an RPO of no more than 15 minutes. Measured recovery steps (18-minute DNS cutover, 25-minute application rebuild, and under 5 minutes to bring the DR DB online) can complete within 60 minutes, so the RTO appears achievable. However, asynchronous log shipping every 20 minutes allows up to 20 minutes of committed transactions to be lost, which exceeds the 15-minute data-loss limit, so the overall availability commitment is not met.
Why not B: This is tempting because candidates often sum failover components and assume the timeline is too long. In this fact pattern 18 + 25 + under 5 is still less than 60 minutes, so the restoration-time requirement is likely met; the clearer failure is the replication interval versus the 15-minute allowable data loss.
Why not C: Backups and restore testing are strong resilience indicators, but they do not eliminate near-real-time data-loss caused by asynchronous log shipping every 20 minutes. Nightly backups are far too coarse to guarantee no more than 15 minutes of lost committed transactions.
Why not D: This distractor appeals to those who separate availability from integrity, but the entity's stated availability commitment explicitly includes a limit on allowable data loss. Therefore meeting the data-loss threshold is part of satisfying the availability commitment.
Question 3
Hint
Compare the replication mode and failover time to the stated RPO/RTO objectives; distinguish availability objectives (RTO/RPO) from performance SLAs.
Answer D. Because the database is synchronously mirrored, committed transactions were present in DC‑West at failover, so the RPO = 0 objective was preserved. Automatic failover completed in 12 minutes, which is within the 30‑minute RTO, so the availability objective was met. The doubled response times indicate a performance SLA breach, which is a degradation in service quality rather than an availability failure.
Why not A: Tempting because losing a primary site sounds like a disaster, but DC‑West continued processing with no data loss and failover completed within the RTO, so production remained available rather than being entirely unavailable.
Why not B: Tempting because traffic resumed and processing continued, but the stem states response times doubled and the performance SLA was violated, so not all production SLAs were being met.
Why not C: Tempting if one assumes asynchronous replication, but the stem specifies synchronous mirroring and that DC‑West contained all committed transactions at failover, so the RPO = 0 objective was not violated.
Question 4
Hint
Ask which risk is controlled by replication frequency (the data‑loss window/RPO) versus which are controlled by failover design (downtime/RTO).
Answer A. Replication frequency defines the recovery‑point objective (RPO), the maximum age of data that can be recovered. Because the DR was last updated at 2:00 AM the previous day, a failure at 1:00 AM leaves nearly a day's worth of transactions missing from the DR, so those orders could be lost. Extended downtime and recovery effort are driven by the lack of hot standby and manual failover (RTO), not the replication interval.
Why not B: This is tempting because the stem explicitly notes manual failovers take about 12 hours, which causes downtime and lost sales. However, those effects stem from failover design (RTO) and the absence of a hot standby, not the once‑daily replication schedule.
Why not C: Tempting because missing transactions can create reconciliation work. But reconciliation is a secondary operational consequence of lost recent data; the most direct consequence of infrequent async replication is the data‑loss window (RPO) itself, actual missing transactions.
Why not D: This distractor plays on real concerns about divergence in async setups and can complicate recovery. However, asynchronous replication scheduled once daily primarily defines how much recent data is absent (RPO); conflicts or complex state reconciliation depend on system specifics and are not the most direct consequence of the once‑daily policy.
Question 5
Hint
Compare the old 20-24 hour recovery and data-loss window from nightly backups with the near-real-time replication and automated failover when assessing changes to RTO and RPO.
Answer A. Moving from nightly backups and a cold standby (20-24 hour RTO and up-to-a-night RPO) to near-real-time replication with a hot standby and automated failover typically cuts recovery time from many hours to minutes/seconds and narrows the data-loss window to a small replication lag. Although asynchronous replication is not always zero-loss, its RPO is far smaller than a full nightly period. Automated failover further shortens RTO, so overall availability usually improves materially.
Why not B: This is tempting because asynchronous replication can lag and might miss the most recent transactions at failover. However, compared with nightly backups (which permit up to a full day's loss), near-real-time replication reduces the potential data-loss window, so RPO improves rather than worsens.
Why not C: This distractor relies on assuming failover is manual. The stem explicitly states automated failover to the hot standby, so recovery time is shortened relative to manual cold-site provisioning, RTO is reduced.
Why not D: Increased complexity and network dependence are valid operational risks, making this choice plausible. But given the concrete improvements described (near-real-time replication to a geographically separate hot standby and automated failover), the net effect on RTO/RPO and availability is typically positive; those risks are usually managed and do not outweigh the availability gains in the scenario.
Question 6
Hint
Ask whether 'asynchronous' guarantees the standby has every committed write at the instant the primary fails; consider the replication lag provided.
Answer B. Asynchronous replication allows the primary to acknowledge commits before those writes reach the standby. With automated failover, availability (RTO) is typically low, the standby can serve quickly, but any commits still in transit at the time of failure will not be present on the standby, producing potential data loss up to the observed replication lag (RPO > 0, up to ~30 seconds).
Why not A: This is tempting because automated failover implies fast recovery and the standby is intended to continue service. It is wrong because asynchronous replication does not guarantee the standby has received every recently committed transaction, commits in the last seconds before failure can be absent at the standby.
Why not C: This distractor is plausible since some systems or policies can delay promotion to avoid data loss. In the given scenario, however, the standby is configured for automated failover and already holds replicated data; asynchronous replication normally means missing recent commits, not a requirement for full restores or long resynchronization that causes an extended outage.
Why not D: This tempts candidates who conflate async replication with multi‑master eventual consistency. It is incorrect here because the system is single‑primary (only one writer), so the usual problem on failover is missing recent commits at the standby, not conflicting writes that require reconciliation.
Common questions
What is the difference between RTO and RPO?
RTO limits how long recovery may take. RPO limits how much recent transaction data may be lost, measured as a time window.
Does synchronous replication meet both RTO and RPO?
Durably storing commits at both sites supports RPO = 0. Recovery time still needs a separate comparison with RTO.
Can asynchronous replication meet a one-minute RPO?
Yes, if its maximum lag is no more than one minute. A five-minute maximum lag cannot guarantee that target, even with automatic failover.
Watch it solved
A full CPA ISC task-based simulation on System availability, worked step by step.
System Availability & RTO/RPO | CPA ISC Simulation (TBS) on YouTube
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 System availability
- When to Trigger Failover: RTO and RPO
- Choose Disaster Recovery Architecture by RTO and RPO
- System Availability: Redundancy and Automatic Failover
- How to Calculate SLA Uptime at 99.95% Availability
- System Availability: Why Backups Do Not Prove Recovery
- Availability vs. Security, Confidentiality, and Integrity
- Availability Calculation: MTBF, MTTR, and Redundancy
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.