Understanding Reliability Metrics: From Time‑Line to Series‑Parallel Calculations

The goal of this lesson is to distinguish reliability, availability and maintainability, to read and compute MTTF, MTTR and MTBF, and to apply the correct formulas for series, parallel and mixed‑system reliability calculations.

Reliability, Availability and Maintainability

Reliability answers the question “Can the system run without failure for a specified time under specified conditions?” Example: a registration service runs 240 hours with a 99.9 % probability of no failure.

Availability measures the proportion of time the system is serviceable during an observation period. Example: 8 760 h per year minus 8.76 h of downtime yields 99.9 % availability.

Maintainability asks whether a failure can be diagnosed, repaired and restored quickly. Common improvements include logging, tracing, health checks, automatic failover, modular design, automated deployment and backup drills. Better maintainability reduces MTTR, which in turn raises availability.

Reliability focuses on “how long it stays up”, availability on “what fraction of time it is up”, and maintainability on “how fast it can be restored”.

Key Time Metrics

Definitions from the official tutorial:

MTTF (Mean Time To Failure): average time a component works before it fails.

MTTR (Mean Time To Repair/Restoration): average time from failure detection to service restoration.

MTBF (Mean Time Between Failures): average interval between two consecutive failures, equal to MTTF + MTTR in the official Chapter 9 definition.

Example calculations:

MTTF = (900 h + 1100 h + 1000 h) / 3 = 1000 h
MTTR = (8 min + 12 min + 10 min) / 3 = 10 min
MTBF = MTTF + MTTR = 1000 h + 0.1667 h ≈ 1000.17 h

Availability Formula

The safe semantic formula is:

A = Average Normal Service Time ÷ (Average Normal Service Time + Average Downtime)

Depending on how the problem defines MTBF, you substitute either MTTF or MTBF for the “normal service time”. The three equivalent forms are:

A = MTTF ÷ (MTTF + MTTR)
A = MTBF ÷ (MTBF + MTTR)   // when MTBF is defined as average uptime
A = MTTF ÷ MTBF            // when MTBF = MTTF + MTTR

Series (串联) Systems

All components must be operational for the system to succeed. Assuming independence: R_series = R1 × R2 × … × Rn Example with three components (gateway 0.99, registration 0.98, database 0.97):

R_series = 0.99 × 0.98 × 0.97 = 0.941094 ≈ 94.11 %

When identical components are repeated, the formula becomes R_series = Rⁿ.

Parallel (并联) Systems

The system succeeds if at least one component works. Compute the probability that *all* components fail and subtract from 1:

R_parallel = 1 - (1‑R1) × (1‑R2) × … × (1‑Rn)

Example with two gateways each 0.90:

Failure probability per gateway = 0.1
All fail = 0.1 × 0.1 = 0.01
R_parallel = 1 - 0.01 = 0.99 (99 %)

With different reliabilities (0.95 and 0.90):

R_parallel = 1 - (0.05 × 0.10) = 0.995 (99.5 %)

Mixed (混联) Systems

Break the diagram into the smallest identifiable series or parallel groups, compute each group's equivalent reliability, replace the group with a single equivalent component, and repeat until a single value remains.

Medical‑platform example:

Gateway parallel: R_gate = 1 - (1‑0.90)² = 0.99
Database parallel: R_db = 1 - (1‑0.95)² = 0.9975
Overall = R_gate × R_registration × R_db
        = 0.99 × 0.98 × 0.9975 = 0.9677745 ≈ 96.78 %

If the registration service also had two 0.98 instances, its parallel reliability becomes 0.9996, raising the overall reliability to about 98.71 %.

Why Real‑World Redundancy May Fall Short

The ideal formulas assume independent failures, flawless failover, and no shared resources. In practice, common‑mode failures such as shared power, network switches, common software bugs, or a single load‑balancer can drastically reduce the effective reliability. Example:

Two servers each 0.99 in parallel → 0.9999
Shared switch reliability 0.98 → Overall = 0.9999 × 0.98 = 0.979902 (≈97.99 %)

Thus, redundancy must be designed with fault‑domain isolation, independent power, network paths, and regular failover drills.

Common Mistakes to Avoid

Treating MTBF always as “average uptime” – the definition varies across sources.

Assuming high reliability automatically yields high availability – frequent short failures can still lower availability.

Adding reliabilities of parallel components (0.9 + 0.9 = 1.8) – probabilities cannot exceed 1.

Believing any extra replica improves reliability without checking independence and failover correctness.

Mis‑interpreting 99.9 % availability as “0.1 h downtime per year” – it actually means 0.1 % of the year, i.e., about 8.76 h.

Quick Review (Three‑Minute Recap)

Reliability = no‑failure probability over a time window.

Availability = serviceable‑time proportion.

Maintainability = ease of repair (affects MTTR).

MTTF, MTTR, MTBF definitions and why MTBF has two conventions.

Use the semantic availability formula to avoid MTBF ambiguity.

99.9 % yearly uptime ≈ 8.76 h of allowed downtime.

R(t) = 1 – F(t) expresses reliability as the complement of failure probability.

Series systems require all components; multiply independent reliabilities.

Parallel systems succeed if any component works; compute 1 minus the product of failure probabilities.

Mixed systems are reduced step‑by‑step by evaluating inner series/parallel groups.

Real‑world dual‑node setups may be limited by shared fault domains.