raid planner

What you actually get, and what it costs you to rebuild

Usable capacity in both TB and TiB, how many failures the array survives, how long a resilver takes, and the odds of hitting an unrecoverable read error while it runs. That last number is the one that kills arrays.

prefill from the index

RAID level

drives in the array

capacity per drive

TB

decimal TB, as soldprice per drive

$

optional — for cost per usable TBtarget pool fill

%

ZFS slows badly past ~80%rebuild speed

MB/s

100 MB/s is typical for a busy arraydrive URE rate

usable capacity

86.4 TB

78.6 TiB as the OS reports it · at 80% fill

raw purchased

144.0 TB

25% lost to redundancy

drive failures survived

2

rebuild time

2.1 days

at 100 MB/s sustained

URE risk per rebuild

57.9%

reads 108.0 TB to resilver

cost per usable TB

—

set a drive price

RAIDZ2 (ZFS)

The homelab default. Two-drive tolerance plus scrubbing.

how this is calculated

Usable capacity is data drives × capacity × your fill target. Vendors sell decimal terabytes (10¹² bytes); your OS reports TiB (2⁴⁰). That is a 9% haircut before redundancy, and it is why a "20TB" drive shows up as 18.2TiB.

URE risk is 1 − e^−bits/rate, where bits is everything the array must read to resilver one failed drive. A consumer drive rated 1 in 10¹⁴ is roughly one error per 12.5TB read — which is why wide RAID 5 arrays of large consumer drives are a coin flip during rebuild. Enterprise nearline is rated 10¹⁵, an order of magnitude better.

Rebuild time assumes sustained sequential throughput. A pool serving live traffic will be considerably slower, and slower rebuilds mean a longer window in which a second failure takes everything.

intellectual property

This calculator — its model, parameters, interface and written methodology — is © 2026 Ugly Pumpkins LLC. All rights reserved. Proprietary. Free to use here; not licensed for reproduction, embedding or derivative works. Enquiries: [email protected]