ISC · Information Systems and Data Management · 6 practice questions
Availability Calculation: MTBF, MTTR, and Redundancy
For independent components, multiply downtime probabilities within a redundant tier, then multiply availability across required tiers. Below, a worked calculation turns MTBF and MTTR into server availability and includes a shared switch.
Try one first
Hint
Compute per-server steady-state availability using A = MTBF/(MTBF+MTTR), then apply the parallel-redundancy formula (1 − (1 − A)^N) to see which option meets the 99.95% target.
Answer A. Single-server availability = 4000/(4000+40) = 4000/4040 ≈ 0.990099 (≈99.01%). With two identical servers and instantaneous automatic failover, A_system = 1 − (1 − 0.990099)^2 ≈ 0.999902 (≈99.9902%), which exceeds the 99.95% SLA. Therefore the redundant active-passive pair meets the SLA under the stated assumptions.
Why not B: Reducing MTTR helps but is insufficient here: A = 4000/(4000+8) ≈ 0.998004 (≈99.80%), which remains below the 99.95% SLA, so this change alone would not meet the requirement.
Why not C: Higher MTBF improves availability, but A = 40000/(40000+40) ≈ 0.999001 (≈99.9001%), which is still below 99.95%, so this alone would not meet the SLA.
Why not D: Nightly backups and manual switchover protect data but are not instantaneous automatic failover; human-driven recovery typically takes many hours and cannot be assumed to achieve 99.95% availability under the problem's steady-state assumptions.
Worked example
A company's application uses two identical servers with automatic, instantaneous failover. Each server has MTBF of 900 hours and MTTR of 100 hours. Both share one switch with 98% availability. Assume independent failures. The application works only when the switch and at least one server are available. Calculate overall availability.
| 1 | Find each server's availability900/(900 + 100) = 0.90 | 90% |
| 2 | Find each server's downtime probability1 − 0.90 = 0.10 | 10% |
| 3 | Combine the redundant servers1 − (0.10 × 0.10) = 0.99 | 99% |
| 4 | Include the required shared switch0.99 × 0.98 = 0.9702 | 97.02% |
The application's overall availability is 97.02%.
Check: Downtime is 2% when the switch is down plus 98% × 1% when the switch works but both servers are down: 2.98%, leaving 97.02% availability.
Key points
- A redundant pair works when either server is available; it fails only when both are down.
- The parallel formula assumes independent failures and instantaneous failover.
- A shared switch remains a required dependency, even when the servers are redundant.
How the exam traps you
- Multiply both server availabilities and the switch: 90% × 90% × 98% = 79.38%. The servers are alternatives, not required series components. Combine their downtime probabilities first.
- Stop at the redundant servers' 99% availability. Include the shared switch: 99% × 98% = 97.02%.
Question 2
Hint
Calculate current system availability (product of the three availabilities). For a tier with availability a, a two‑unit parallel configuration has availability 1 − (1 − a)^2; compute the new system availability for each possible duplication and compare the absolute increases.
Answer A. Initial overall availability = 0.998 × 0.999 × 0.9995 = 0.996503499 (≈99.65035%). Duplicating the ISP gives Internet availability 1 − (1 − 0.998)^2 = 0.999996, so new overall ≈ 0.999996 × 0.999 × 0.9995 = 0.998496505998 (≈99.84965%), an absolute increase ≈0.001993007 (≈0.19930 percentage points). Duplicating the application or database yields smaller absolute increases (≈+0.000996503 and ≈+0.000498252, i.e., ≈+0.09965 and ≈+0.04983 percentage points), so adding the second ISP produces the largest gain.
Why not B: Tempting because the application tier is critical, but its initial unavailability (0.10%) is smaller than the Internet's (0.20%). Duplicating the app reduces overall unavailability by about 0.0009965 (≈0.09965 percentage points), which is less than the improvement from duplicating the ISP.
Why not C: Tempting because the database is important, but the DB already has the highest single-unit availability (99.95%), so its unavailability is smallest. Duplicating it yields only a marginal absolute increase (≈0.00049825 or ≈0.04983 percentage points).
Why not D: Incorrect: the absolute improvement depends on each tier's initial unavailability. Because the Internet link has the largest initial unavailability, duplicating it produces the largest absolute increase rather than all tiers producing the same change.
Question 3
Hint
Compute parallel availability as 1 − Π(failure probabilities) for redundant components, multiply series components, and then identify which single series element has the lowest availability (the biggest bottleneck).
Answer B. Compute web-tier availability: 1 − (1 − 0.995)^2 = 0.999975. Baseline overall availability = 0.999 * 0.999975 * 0.998 * 0.996 ≈ 0.99299 (≈99.30%). Making the database redundant yields DB availability 1 − (1 − 0.996)^2 = 0.999984 and raises overall availability to ≈0.99696 (≈99.70%). This single change gives the largest absolute improvement because the nonredundant database was the lowest-availability series component.
Why not A: Redundancy for the load balancer greatly increases its own availability (≈99.9% → ≈99.9999%), but the application still contains other series components (database, switch) with lower availabilities; substituting the LB's improved availability yields a smaller net increase than fixing the database bottleneck.
Why not C: Adding a third web server slightly raises the web-tier availability (from ≈0.999975 to ≈0.9999999), but the web tier was already very reliable and the incremental gain is negligible compared with improving the much-lower-availability database.
Why not D: Making the switch redundant increases its availability (≈0.998 → ≈0.999996), which helps, but the original switch availability is higher than the database's 99.6%, so improving the switch yields a smaller overall increase than removing the database single point.
Question 4
Hint
Compute each component's availability (MTBF/(MTBF+MTTR)), treat the two servers as parallel (cluster availability = 1 − (1−A)^2), then multiply series component availabilities and compare to 99.95%.
Answer B. Compute component availabilities: each server A = 2000/(2000+10) ≈ 0.995025 (99.5025%). Two-server cluster (parallel) availability = 1 − (1−A)^2 = 40400/40401 ≈ 0.99997525 (99.9975%). SAN availability = 50000/(50000+5) = 10000/10001 ≈ 0.9999 (99.99%); router = 0.995 (99.5%). Multiply the series: 0.99997525 × 0.9999 × 0.995 ≈ 0.994876 (≈99.49%), which is below the 99.95% SLA, therefore the architecture does not meet the requirement, and the single WAN router is the primary limiting factor.
Why not A: Tempting because the server cluster and SAN individually appear nearly fail‑safe, so omitting or downplaying the router can yield a high figure (≈99.98%). However, including the 99.5% router in series gives an overall availability of ≈99.49%, which is below the 99.95% SLA.
Why not C: Tempting because the SAN has a much higher MTBF and redundancy normally helps availability, so improving storage feels like the next fix. But SAN redundancy alone does not change the router's 99.5% single‑point availability in series; the end‑to‑end product remains below 99.95%.
Why not D: Tempting because reducing MTTR greatly increases server availability and the cluster becomes effectively near 100%, so it seems the SLA could be met. However, even with much higher server availability the 99.5% router still caps the overall availability (final product stays ≈99.49%), so the SLA remains unmet.
Question 5
Hint
Calculate baseline and post-change availability using MTBF/(MTBF+MTTR) for both options and compare the absolute (percentage-point) increases rather than relative percent change.
Answer A. Compute availability for each case. Baseline availability = 1000 / (1000 + 10) = 1000/1010 ≈ 99.01%. If MTBF increases to 1,500: 1500 / (1500 + 10) = 1500/1510 ≈ 99.34% (≈ +0.33 percentage points). If MTTR decreases to 2: 1000 / (1000 + 2) = 1000/1002 ≈ 99.80% (≈ +0.79 percentage points). The MTTR reduction yields the larger absolute increase in steady-state availability, so option A is correct.
Why not B: This is tempting because higher MTBF reduces failure frequency, but numerical calculation shows the availability gain from increasing MTBF (≈ +0.33 percentage points) is smaller than the gain from reducing MTTR to 2 hours (≈ +0.79 percentage points).
Why not C: Some candidates assume proportional changes have equal effects, but availability depends on MTBF/(MTBF+MTTR). Explicit calculation shows the two proposed changes produce different absolute (percentage-point) gains (≈0.33 vs ≈0.79).
Why not D: This confuses availability with recovery objectives. RPO/RTO and backups address data loss and recovery timing, not steady-state uptime, which is determined here by MTBF and MTTR (and any redundancy/failover).
Question 6
Hint
First compute the combined availability of the two servers using the redundancy formula 1 - (1 - p)^n; then include any nonredundant components by multiplication.
Answer D. First compute the combined availability of the two redundant servers: 1 - (1 - 0.98)^2 = 0.9996 (99.96%). Then include the single points of failure by multiplication: 0.9996 × 0.999 × 0.999 ≈ 0.9976, or about 99.76%. This accounts for server redundancy and the shared network switch and power feed.
Why not A: This result comes from multiplying all component availabilities directly (0.98 × 0.98 × 0.999 × 0.999), which incorrectly assumes both servers must be working simultaneously. Redundant servers use the complement rule (1 - (1 - p)^n), not simple multiplication.
Why not B: This is the availability of the two servers considered alone (1 - (1 - 0.98)^2 = 99.96%). It neglects the shared network switch and power feed, which are single points of failure that reduce overall availability when included.
Why not C: This treats the system as if only one server provides service (ignoring the passive failover). Because of redundancy, the combined server availability should be higher than a single server's 98% before accounting for shared components.
Common questions
How do you calculate availability from MTBF and MTTR?
Divide MTBF by MTBF plus MTTR. Here, 900/(900 + 100) = 90% availability for each server.
How do you calculate availability for two redundant servers?
With independent failures and instantaneous failover, subtract the probability both are down from 1. Here, 1 − (0.10 × 0.10) = 99%.
How does a shared switch affect redundant server availability?
The switch must work even if a server is available. Multiply the pair's 99% availability by the switch's 98% availability to get 97.02%.
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
- RTO vs RPO: Synchronous vs Asynchronous Replication
- System Availability: Why Backups Do Not Prove Recovery
- Availability vs. Security, Confidentiality, and Integrity
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.