Starlight Tools
Three 99.9% components in series
0.999³ = 0.997002999
Combined availability is about 99.7003%, or approximately 1 day 2 hours of equivalent downtime per 365-day year.
Infrastructure Availability Calculator - Series, Parallel and Redundant Components
Model series dependencies, parallel service paths, and N+1, 2N or general k-of-n component groups. See the combined availability, equivalent downtime, and the reliability of every block.
Availability model
Architecture inputs
Each row is one block of identical components. Enter how many units are installed (n) and how many must work (k).
Redundancy inside each block is controlled by k and n below.
| Block | Unit availability (%) | Installed n | Required k | Action |
|---|
Results
Ready to calculate
Use a preset or describe your architecture, then calculate.
System availability
—
—
System unavailability
Modeled unavailable fraction
Equivalent annual downtime
—
365-day planning year
Largest series risk block
—
—
| Period | Equivalent downtime |
|---|---|
| Day | — |
| Week | — |
| 30-day month | — |
| 365-day year | — |
| Block | Configuration | Block availability | Unavailability | Equivalent downtime/year | Failure tolerance |
|---|---|---|---|---|---|
| No calculation yet. | |||||
Advertisement
Advertisement
How the availability model works
Series blocks: A_system = A₁ × A₂ × … × Aₘ
Parallel blocks: A_system = 1 − [(1 − A₁) × (1 − A₂) × … × (1 − Aₘ)]
Identical k-of-n block: A_block = Σ C(n,i) × Aᶦ × (1 − A)ⁿ⁻ᶦ, from i = k through n.
Equivalent downtime: (1 − A_system) × period length
This calculator applies Boolean reliability-block formulas to availability probabilities measured over a common period. A row with n = 1, k = 1 is a single required component. A 1-of-2 row represents two independent active components where either can carry service. A 2-of-3 row represents a quorum or N+1 group that needs any two units. A 2-of-4 row can represent 2N only when any two units truly have enough capacity and can take over.
Shared-fate warning: mathematical redundancy helps only when failures are sufficiently independent. Power, cooling, network, identity, deployment, control-plane and human dependencies can make nominally separate units fail together. Use the shared-dependency field for one known common block, but treat this calculator as a simplified reliability-block model.
Series, parallel and redundant examples
Two independent 99.9% components in parallel
1 − (0.001²) = 0.999999
A 1-of-2 pair is 99.9999% available in the ideal independent model, about 31.5 seconds of equivalent downtime per year.
Two required out of three at 99.9%
C(3,2) × 0.999² × 0.001 + 0.999³
The modeled 2-of-3 block is 99.9997002% available before shared dependencies and failover risks are added.
Assumptions, limits and sources
Reviewed: 28 July 2026. Inputs are processed only in your browser by this page's calculator code and are not uploaded by the tool.
- Availability inputs must describe the same operating period, service boundary and definition of “available.” Do not mix incompatible SLAs, host reliability figures and request-success SLIs without a justified mapping.
- Units within each k-of-n row are assumed identical and independent. Blocks are also assumed independent except for the optional shared dependency.
- Failover detection, switching, quorum, data consistency and capacity are assumed to work. If they have separate availability, enter them as required series blocks.
- Equivalent downtime distributes the unavailable fraction across a period. It does not estimate the number, timing or duration distribution of incidents.
- Rounded supplier SLAs are commitments, not necessarily measured component probabilities. Correlation can dominate very high “nines.” Validate the architecture with dependency mapping, failure testing and observed service-level indicators.
References
- NIST/SEMATECH: Series model — series reliability and independence assumptions.
- NIST/SEMATECH: R out of N model — the identical-component k-of-n binomial model.
- NIST/SEMATECH: Complex systems — reducing combinations of series, parallel and r-of-n blocks.
- AWS Well-Architected: deploy workloads to multiple locations — redundancy benefits require components to operate independently.
Frequently asked questions
How is series availability calculated?
When every independent block must work, multiply their availabilities. A series chain can therefore be less available than any one component.
How is parallel availability calculated?
When any one independent block can provide service, multiply their unavailabilities and subtract the result from one. This assumes each path can carry the required service and does not share the same failure.
What does k-of-n redundancy mean?
A k-of-n block has n installed units and remains available while at least k work. N+1 is commonly k required out of k+1 installed. 2N can be k required out of 2k installed when any k units can carry the full demand.
Why can calculated redundancy overstate real availability?
Independent arithmetic does not capture common power, networking, software, control-plane, maintenance, overload or human failures. Those shared risks often limit real-world improvement.
Can I model active-passive or cold-standby failover?
Only approximately. A k-of-n block assumes eligible units are available when needed. Include the failover mechanism as a required block or fold tested transfer performance into the inputs. Cold-standby lifetime analysis needs a different model with failure rates and switching behavior.
Is equivalent downtime a prediction or an SLA?
No. It is simply the unavailable fraction multiplied by a time window. Contracts may exclude events, use request-based measures, or define service periods differently.