Data Recovery Case File · Solid State & Flash · Both Halves or Nothing

RAID 0 SSD Recovery: Both Halves or Nothing

His enquiry supplied exactly the detail that decides the whole approach, and it is a configuration most people never realise their laptop had. Storage from a Sony VAIO laptop: "they are 2 separate SSDs raided to make 256GB." The material he needs is photographic — "mainly images, either CR2 or JPG." That one line about two drives striped together changes everything, because a striped pair is not two drives holding half the files each. It is two drives that each hold half of every file, alternating in chunks, and the consequence is stark: a single module examined on its own contains no complete photograph at all. This page explains what that means for imaging, for carving, and for the honest odds — and why the pair must never be separated in anyone's thinking, even when they are physically apart.

The decode: what striping does to a file, and why carving one module fails

How a stripe stores a photograph: striping splits data into fixed-size chunks and writes them alternately across the members — first chunk to module one, second to module two, third to module one, and so on. The purpose is speed: two devices reading simultaneously deliver data faster than one. The consequence for recovery is that a single photograph of any size is physically divided between the two modules. Neither one holds a complete file; each holds alternating slices of every file on the machine.

Why examining one module produces nonsense: this is where cases like his most often go wrong, and it is worth being concrete. If a recovery tool is pointed at one module alone, it will find the beginnings of images — the signatures that mark the start of a CR2 or a JPEG are at the front of the file, so they are genuinely present on whichever module received that first chunk. It will then dutifully carve "files" that consist of the first slice of one photograph followed immediately by a slice from somewhere else entirely. The results look plausible in a file listing, and they open as an image that renders correctly for a fraction of its height before turning to noise. Owners see that, conclude the data is corrupted beyond saving, and stop. It is not corrupted; it has been read at the wrong level.

What has to happen first: the array must be reassembled before any file-level work begins. Each module is brought up on a generation-appropriate interface — these are proprietary blades, not standard drives — and imaged individually. Then the stripe parameters are determined from the images themselves: which module came first, how large each chunk is, and where the data area begins. Those are established by analysis rather than assumed, because getting any of them wrong produces exactly the plausible-looking nonsense described above. Only once the two images are logically recombined into the single volume the laptop saw does the filesystem make sense — and only then does carving return whole photographs.

The honest limit of striping: stated plainly, because it is the trade-off this configuration makes. A stripe has no redundancy whatsoever. If one module is unreadable, the volume is missing every alternate chunk of every file, and what can be recovered is limited to material small enough to have fitted within surviving chunks — which for full-resolution camera RAW files is very little. Both members matter equally, and both must be preserved and delivered together. It is the reason this page's advice is really one instruction: never treat one half of a striped pair as a drive in its own right.

On the bench

Both modules were brought up on generation-appropriate proprietary interfaces rather than generic adapters, and each was imaged individually and completely before any interpretation was attempted — the pair frozen safe as copies. The stripe parameters were then derived from the images: member order, chunk size and data offset established by analysis, with candidate reconstructions tested against the filesystem's own structures until the volume resolved coherently rather than approximately. With the array reassembled logically from the two copies, format-aware carving ran for camera RAW and JPEG together, recovering full-resolution originals rather than the previews embedded within them, and every image was validated by rendering at full size. The set was ordered by embedded capture date and delivered on fresh media.

The outcome

Both modules imaged, the stripe reconstructed from the copies, and the photographs carved and validated from the reassembled volume. Free assessment, one fixed written figure including VAT; where a chip has to be removed, 50% of parts and labour is payable upfront with the balance only on success — otherwise no recovery, no fee. The decode, for anyone whose laptop turns out to have had two drives inside it: striping divides every file across both members in alternating chunks, so neither module holds a complete photograph and neither is a drive in its own right; carving a single member returns files that look plausible and render as a slice of one image followed by noise, which is the commonest reason people wrongly conclude their data is destroyed; the stripe parameters must be reconstructed by analysis before any file work; and because a stripe carries no redundancy, both halves matter equally and must always travel together.

Our case files are drawn from genuine enquiries received by our laboratory over the past ten years, anonymised to protect client confidentiality. Each one describes the diagnostic and recovery procedure our engineers apply to that fault, using the equipment listed.