People arrive here with one of two questions. Should my next array be RAID 10 or RAID 5? or my RAID 10 has just dropped a drive — how worried should I be? They share a root answer, because both come down to one thing: the way RAID 10 keeps its copies, and the cold arithmetic of what happens when the next disk goes.
RAID 10 pairs its drives into mirrors — each pair an identical twin set — then stripes your data across those pairs for speed. Quick, because the pairs share the load; resilient, because every pair is a backup of itself.
RAID 5 lays your data down alongside a calculated backstop — parity — spread across every drive: buy N disks, keep the capacity of N−1, and any single failure is survivable. RAID 10 throws out the maths and simply keeps two of everything. You give up half your capacity, but writes are quick because nothing has to be computed, and replacing a dead disk is a plain copy from its twin — a rebuild that runs in hours and leans on one surviving drive, rather than a parity rebuild that grinds every disk in the set for however long it takes.
That headline you see everywhere — ‘RAID 10 survives multiple failures’ — comes with fine print about where. Two failures in separate pairs, no problem. A second failure inside a pair that has already lost a drive takes that pair’s share of every file down with it, because those two twins were the only copies there were. And reality loads the dice: a pair is usually two identical drives from one batch, fitted the same afternoon and worn by exactly the same work — so when one goes, its twin is often not far behind, and the rebuild’s heavy reading is what finishes the job.
The rule that keeps a wounded array savable fits on a Post-it: shut it down, photograph and label every drive with its bay, and change nothing. Done properly, each member is then imaged read-only and the array rebuilt from those images, so the rebuild happens on copies and never on your last surviving originals. That approach is the spine of our RAID recovery work, whether the ‘array’ is a rack server or a two-bay NAS on a shelf.
No — and that mix-up keeps recovery labs busy. RAID 10 guards against exactly one thing: a drive dying. Accidental deletion, ransomware, corruption, a controller going bad, fire, theft — all of those copy themselves instantly and faithfully onto both mirrors. An array keeps you trading through a disk failure; only a separate, disconnected copy keeps you safe from the rest. You want both, for different jobs.
The safety case holds; the speed case shrinks. SSDs are quick enough that striping matters far less than it did with spinning platters, and plenty of builds get the same protection more cheaply from a simple RAID 1 mirror. Where SSD RAID 10 still earns its place is heavy sustained writes — busy databases, virtualisation — where both the raw throughput and the fast, gentle rebuilds are worth paying for.
It can — and that’s how plenty of two-drive disasters begin. A degraded array runs with no protection left on that mirror, while the dead drive’s partner — same model, same age, same mileage — shoulders the load alone. If the data matters, treat ‘degraded’ as an event and not a state: back up now, swap the drive now, and if anything about the rebuild feels shaky, image it first.
Free 48-hour diagnostic on multi-disk arrays — every member imaged read-only, every rebuild run against the copies, a written quote before any work.