NAS Capacity Calculator
Calculate usable NAS capacity from disk count, disk size, RAID efficiency, and filesystem overhead.
Inputs
Fraction of raw capacity retained after RAID (e.g. 75% for RAID 5 across 8 disks, 60% for RAID 6)
Usable Capacity
22.80TB
Raw Capacity
32.00TB
Filesystem Overhead
1.20TB
Overall Efficiency (usable / raw)
71.3%
Step by step
Values used
Number of Disks = 8; Size per Disk = 4 TB; RAID Efficiency = 75; Filesystem Overhead = 5
Usable capacity
usable = disks × disk_size × raid_efficiency% × (1 − filesystem_overhead%)
Usable Capacity
= 22.80 TB
Raw Capacity
= 32.00 TB
Filesystem Overhead
= 1.20 TB
Overall Efficiency (usable / raw)
= 71.3
How it works
NAS usable capacity starts from raw capacity (disk count × per-disk size), applies the RAID level's efficiency factor to account for parity or mirroring, then subtracts filesystem overhead for inode tables, journaling, and reserved blocks that most filesystems (ext4, XFS, ZFS) set aside. The result is a realistic estimate of the space actually available for files, typically 5-15% below the RAID-adjusted figure.
Formula
Usable capacity
usable = disks × disk_size × raid_efficiency% × (1 − filesystem_overhead%)
Frequently Asked Questions
What RAID efficiency should I use for RAID 5 vs RAID 6?
RAID 5 efficiency is (n-1)/n where n is the disk count in the group — 87.5% for 8 disks. RAID 6 is (n-2)/n — 75% for 8 disks. RAID 10 is always 50% regardless of disk count since every disk is mirrored.
Why is filesystem overhead typically higher on NAS than block storage?
NAS exposes a filesystem directly (NFS/SMB shares) rather than raw blocks, so it always carries filesystem structure overhead — inode tables, journal, superblocks — on top of any RAID overhead, whereas some SAN/block deployments let the host filesystem's overhead be counted separately.
Does deduplication or compression change this calculation?
No — this calculates raw usable capacity before any data reduction. Deduplication and compression happen at the data layer on top of usable capacity and their effective ratio depends heavily on data type, so they're modeled separately rather than as a fixed overhead percentage.