RAID Capacity Planner

Reading the result

The usable figure this returns is capacity after parity and before the filesystem takes its share. ZFS, Btrfs and ext4 each reserve a slice for metadata and free-space management, so plan on landing somewhat under the number shown rather than exactly on it. Pools also behave poorly when they approach full — leaving meaningful headroom is a performance decision, not a rounding one.

The parity level is the decision this page exists to inform, and capacity is what shifts it. RAID 5 survives one drive failure; RAID 6 survives two. That sounds like belt-and-braces until you consider what happens during a rebuild: resilvering a large modern drive takes many hours and often more than a day, and for that entire window a single-parity array has no redundancy left. The larger the drives, the longer that window, which is why the standard advice tightens toward RAID 6 and RAIDZ2 as capacity climbs rather than staying fixed.

RAID 10 answers a different question again. It gives up half the raw capacity, which looks expensive next to parity, and buys rebuild behaviour that is a straight copy rather than a reconstruction across every remaining drive. For workloads that cannot tolerate a long degraded window that trade is often the right one, and the calculator shows all three side by side precisely so the cost of each is explicit.

The drives themselves still matter

No parity level rescues an array built on shingled drives. SMR recording overlaps tracks to gain density, and the rewrite behaviour that follows can turn a rebuild into a multi-day event or fail it outright — the one scenario where redundancy is supposed to save you. Verify recording technology before buying rather than after; our model resolver answers it per part number from the manufacturer's own documentation.

One last thing this calculator cannot model: bays you leave empty. Filling every slot on day one means the next capacity increase is a full replacement cycle rather than an addition, and that is usually the more expensive path. Running one or two bays short is often the cheaper long-run configuration even though it looks wasteful on the day.