Source: https://guthlabs.ai/docs/instruments/raid-capacity-fault-tolerance-calculator/mechanism/

instrument mechanism / verified-clean

# RAID Capacity & Fault-Tolerance Calculator mechanism

The inspectable method record for RAID Capacity & Fault-Tolerance Calculator, kept separate from implementation and qualification.

Status: **VERIFIED_CLEAN**Lifecycle: **verified-clean**

## Model scope

The capacity formulas are broadly standard under equal-drive assumptions, but the specification needs explicit validity constraints, an overhead formula, and a defined RAID10 tolerance policy. The honest-label warning is appropriate, but it should explicitly state that RAID10 tolerance varies by mirror-pair placement and that displayed capacity excludes vendor and filesystem overhead unless those are modeled.

- RAID0 capacity N×S is correct for equal-sized drives, but RAID0 tolerates zero drive failures and requires a valid drive-count constraint.

- RAID1 usable capacity S is correct for a mirrored set of equal-sized drives, but the formula omits the required minimum drive count and does not define tolerated failures for N-way mirrors.

- RAID5 capacity (N-1)×S and one-drive tolerance are correct only for N≥3 equal-sized drives.

- RAID6 capacity (N-2)×S and two-drive tolerance are correct only for N≥4 equal-sized drives.

- RAID10 capacity (N/2)×S is correct only when N is even and drives are arranged as equal-sized mirror pairs; tolerated failures are not a single fixed number. Four drives can tolerate one failure in the worst case and two failures if they occur in different mirror pairs.

- The specification claims an overhead-percentage output but provides no formula. Assuming raw capacity is N×S, overhead should be (raw capacity−usable capacity)/raw capacity×100.

## Units and conventions

- All drives have equal usable capacity S.

- Drive capacity and usable capacity use the same unit; the test vectors use decimal TB, where 1 TB=1,000,000,000,000 bytes.

- Overhead percentage is interpreted relative to raw capacity N×S, excluding filesystem, metadata, hot-spare, vendor-reserved, and thin-provisioning overhead.

- RAID10 drive count is even and represents complete mirror pairs.

- The RAID level is implemented as conventional capacity-only RAID, without vendor-specific reserved space or formatting losses.

## Inspectable vector

One of 4 source-bound review vectors:

````
{
  "inputs": {
    "raid_level": "RAID0",
    "number_of_drives": 4,
    "drive_capacity": 2,
    "capacity_unit": "decimal TB"
  },
  "expected_outputs": {
    "usable_capacity": 8,
    "capacity_unit": "TB",
    "redundancy_overhead_percent": 0,
    "tolerated_simultaneous_failures": 0,
    "rebuild_risk_note": "No redundancy; any single drive failure loses the array. Vendor- and workload-specific rebuild/URE analysis is not performed."
  },
  "working": "Raw capacity = N×S = 4×2 = 8 TB. Usable capacity = 4×2 = 8 TB. Redundancy overhead = (8−8)/8×100 = 0/8×100 = 0%. RAID0 tolerated failures = 0."
}
````

## Failure boundary

**BREAK_CLAIMED** · SELF_FLAGGED

The zero-drive input causes a negative usable-capacity result and an undefined redundancy-overhead division by zero, so the instrument must reject it. The unit ambiguity can also produce materially different byte capacities unless TB versus TiB is made explicit.

## Other records

[01**Overview**](<https://guthlabs.ai/docs/instruments/raid-capacity-fault-tolerance-calculator/>)[03**Implementation**](<https://guthlabs.ai/docs/instruments/raid-capacity-fault-tolerance-calculator/implementation/>)[04**Qualification**](<https://guthlabs.ai/docs/instruments/raid-capacity-fault-tolerance-calculator/qualification/>)

## Sources and evidence

These pointers bind the page to evidence. They do not grant authority or collapse verification into deployment.

- `VERIFIED_LEDGER.json`

- `evidence-tranche4/tranche4-results.json`

- `BOUNDARY_HAZARD_CATALOG.json`