By manufacturer
Dell PowerEdge HPE ProLiant Lenovo ThinkSystem Fujitsu Primergy Supermicro IBM System x / Power Acer AltosRAID 0: no redundancy. Recovery almost always possible.
RAID 0 spreads data across two or more disks (striping) with no redundancy: a single failed disk takes down the entire array. Recovery means analysing the surviving disks, where necessary physical work on the failed one (donor heads, PCB transplant), and logical reconstruction of the stripe.
Three mistakes that end the recovery.
- Do not write to the surviving disks. A failed RAID 0 means the volume will not mount. Forcing writes to it (chkdsk, a test mkfs, a "we'll recover it later anyway" format) wipes out the residual metadata the recovery depends on.
- Do not open the failed disk outside a clean room. Spinning up a drive whose heads are already contaminated finishes it off.
- Do not replace disks and change their physical order without documenting it first. Stripe order is critical to rebuilding the data.
Analysis, cloning, logical reconstruction.
- Bit-by-bit cloning of the surviving disks in the lab (we work on the copies, never on the originals).
- Recovery of the failed disk where feasible: same model/firmware donor, PCB or head transplant in a clean room, aggressive reads with dedicated tooling (PC-3000, DDR).
- Logical stripe reconstruction: chunk size, disk order and start offset, derived from the controller metadata or from filesystem analysis.
- Data extraction: returned on new media supplied by the customer.
The questions we get most often.
If one of the RAID 0 disks is physically dead, is the data lost?
Not necessarily. On spinning disks with physical damage we can often recover through component transplants (PCB, heads) in a clean room. On electrically damaged SSDs it is harder. The preliminary diagnosis tells you how much can realistically be recovered.
Is RAID 0 really that risky? Why use it at all?
For raw performance: on a fixed budget, RAID 0 delivers the highest throughput. It only makes sense for transient workloads (rendering scratch storage, compute swap) where the data is volatile by design and a primary source always exists. For data that has value, RAID 0 is almost never the right choice.