A striped array holds no redundancy at all, so one failed member takes the entire volume even though every other disk is healthy. That makes it the most alarming array to lose and, very often, one of the more recoverable, because the data is intact and merely interleaved. This page covers stripe size, disk order and offset, what to refuse when a controller offers to help, and what the array band costs.
Striped sets are the array band: from £500 + VAT, rising with the member count, because every disk is imaged separately before the volume can be rebuilt. The diagnostic is free, and the healthy members still hold their share of every file.
A RAID 0 fails all at once, because there is nothing in it to absorb a fault. What matters is which of these happened, since the answer decides whether the work is a straightforward reassembly or something considerably slower.
The classic case, and the reason this page exists. A striped set has no redundancy at all, so a single failed disk removes the volume even though the other disks are perfectly healthy and still hold their share of every file. The healthy members are imaged, the failed one is worked on, and the stripe is put back together from the copies.
A loose cable, a marginal power connector, a controller port that has gone intermittent. The disk is fine; the array simply lost sight of it long enough to be marked out. This is the best possible version of the problem, and the worst thing to do is start re-creating arrays to see whether the volume comes back.
A BIOS reset, a cleared CMOS battery, a motherboard swap, a card moved into another machine. The disks are untouched and the metadata describing the stripe may still be on them. What must not happen is accepting an offer to initialise, create or repair the set, because that writes a new description over the old one.
Under pressure, this feels like the obvious fix: make the array again with the same disks and see whether the data reappears. Whether it cost anything depends entirely on the stripe size and the offset the new array used, and on whether it was told to initialise. It is still worth a diagnostic, because a re-creation that matched the original layout may have cost nothing at all.
All the disks are present and the volume mounts as raw, or as unformatted, or with half the folders missing. That is a file system problem sitting on top of a working array — a bad shutdown, a failed resize, ransomware — and no disk needs opening for it. The array band still applies because the volume must be reassembled before anything can be read.
Two-bay NAS units are frequently shipped in RAID 0 to advertise the bigger number, and Apple RAID and Windows spanned volumes do the same thing in software. Plenty of people discover they had a striped set only at the point one disk failed. The layout is written on the disks either way and can be worked out again.
The name is misleading. RAID stands for a redundant array, and level 0 is the one level with no redundancy in it whatsoever — the zero is there because it protects nothing. A striped set takes your data and cuts it into fixed-size pieces, then writes those pieces to each disk in turn: the first chunk to disk one, the second to disk two, the third back to disk one, and on down the volume. Nothing is duplicated. Nothing is calculated. Every disk holds a different fraction of every large file.
What you get for that is speed and capacity. Two disks read and write at once, so a video editor or a database sees close to double the throughput of a single drive, and two 4TB disks present as one 8TB volume with nothing held back. What you give up is any margin at all. A mirror survives a disk death. RAID 5 survives one. A striped set survives none, and worse, the chance of losing one disk out of several is higher than the chance of losing a single disk, so a RAID 0 is measurably less reliable than the plain drive it replaced.
That combination is why this is the most alarming array to lose. Everything is on it because it was the fast, big volume; nothing is protected because that was the trade; and the failure is total rather than degraded. There is no window in which the volume limps along giving you time to copy things off.
Here is the part that gets lost in the panic. Losing a striped set does not destroy the data on the surviving members. Every one of those disks still holds its own chunks, intact, in the order they were written. The volume is unreadable not because the bytes have gone but because they are interleaved, and the map explaining the interleave lived in the controller rather than in the files.
Reassembly is therefore a matter of working out three things and putting the pieces back in order. The first is the disk order — which member came first, second, third. The second is the stripe size, the length of each chunk, commonly 64KB or 128KB but by no means always. The third is the offset, meaning how far into each disk the striped region actually starts, since controllers reserve a little space at the front for their own metadata.
Those three can be derived from the disks themselves without the controller. Large files that spread across every member leave recognisable patterns at the boundaries, and file system structures land at predictable places once the geometry is right. Get all three correct and the volume appears exactly as it was. Get one wrong and you get a mount that looks plausible and produces corrupt files, which is precisely why this should not be attempted by trial and error on the original disks.
When one member is genuinely dead, that member is imaged first — repaired far enough to be read, not far enough to be used again — and the image takes its place in the reassembly. If a region of it cannot be read at all, the damage is confined to the files whose chunks landed there rather than spread across the whole volume, so a partial result is usually a large partial result.
Do not let anything initialise, create or repair the array. Every controller offers it and the wording is always reassuring. A metadata-only initialise is survivable; a full one writes across the members and is not.
Do not shuffle the disks between ports or bays, and if they have already come out, do not guess at the order when putting them back. The order can be worked out from the data, but it costs time that need not be spent, and a wrong guess combined with a file system repair tool is how a recoverable job becomes an expensive one.
Do not run CHKDSK, fsck or a disk repair utility against a striped volume that is mounting oddly. Those tools assume the storage beneath them is sound and will happily write their conclusions across data that was intact, on every member at once.
Do not keep powering up a member that clicks, grinds or spins up and down. A disk making mechanical noise is losing readable surface every minute it runs, and in a striped set that disk holds a share of every single file.
Any storage holding more than one disk is the array band: from £500 + VAT, rising with the number of members, because each disk is imaged separately before reassembly can start. A two-disk desktop stripe sits at the bottom of that band. A striped set spread over four or six disks costs more, and the figure is fixed in writing after the free diagnostic rather than estimated on the phone.
The rest of the table, so nothing is a surprise: £250 + VAT for a memory card or USB stick; £300 + VAT for one hard drive or SSD; £400 + VAT for a CCTV or DVR recorder disk and for BitLocker or other encrypted volumes; £800 + VAT for forensic investigation with a written report, or £400 + VAT for a binary image with deleted-file extraction and no report. Ransomware on an array is priced as ordinary array work, from £500 + VAT.
Where the disks are all healthy and the problem is the array or the file system, the job is a logical one and carries no fix, no fee. Once a member has physically failed it is not: electronic and mechanical failures and chip-level work are published exclusions, along with DVR and forensic jobs, and physical work takes fifty per cent of the quoted figure upfront.
The diagnostic is free and closes two working days after the disks are booked in. Multi-disk work runs longer than the two to four working days a single drive takes, and the honest reason is arithmetic rather than queueing: three members are three imaging passes, and only when the last one finishes can the geometry be worked out and the volume rebuilt. A two-disk stripe with one dead member is usually the quickest of the array jobs; a six-disk set with a difficult drive in it is the slowest.
Send the member disks only. Not the chassis, not the controller card, not the tower they came out of. Write the bay number on each disk in marker before it is pulled — one, two, three, left to right or top to bottom as the unit presents them — and photograph the front of the machine first. Knowing the order for certain saves working it out from the data, and although it can be worked out, it takes a while.
The single exception is a NAS. A two-bay or four-bay network unit travels whole with its disks left in their bays, which keeps the order intact without anybody having to record it. Pack it so nothing moves and leave the power adaptor at home.
The address is Oxford Data Recovery, John Eccles House, Oxford Science Park, Robert Robinson Avenue, Littlemore, Oxford OX4 4GP. Drop-offs are taken there on weekdays, 9:00am to 5:30pm, and Coventry to Oxford is about fifty-five miles of M40. There is no counter in Coventry and no collection service.
The RAID 0 work that reaches this bench from Warwickshire follows the trades. Editing volumes in video and design studios, where the speed was the whole point. Scratch arrays in engineering offices around Ansty and Whitley running simulation output. Two-bay NAS units in small practices that were bought striped without anybody realising it. Workstations in the motor trade with a fast scratch volume for scan and CAD data. In almost every one of them the volume held the working files, the backup was of last month, and the fast disk was the one nobody thought would stop.
The surviving disks still hold their chunks intact. Getting the volume back is a matter of order, stripe size and offset, worked out on copies.
Parcels are how nearly all of this work arrives, and they come from a good deal further afield than Warwickshire. Padding and a tracked label treat a poorly drive better than a week of being carried about, and a box handed to a courier on a weekday afternoon usually reaches the Oxford bench in the morning.
As a rule the storage comes out and the machine stays where it is. That applies to a laptop, a tower, an iMac and to the recorder sitting under a counter. Taking equipment apart is not something this bench does, and a repair shop will free a drive in a few minutes. Two things go the other way: an external drive stays sealed inside its own case, and a NAS travels as a complete unit with its disks still in their bays. A Fusion Mac is a third case — both of its drives come out and travel together, each one labelled. The single situation nobody can work around is memory soldered flat onto a mainboard, which is how Apple Silicon Macs and a good many slim laptops are built: if the storage will not unbolt, there is no parcel to send.
↓ Print the shipping & booking-in form (PDF)
Put Oxford Data Recovery on the label. From Coventry it is roughly fifty-five miles straight down the M40, about an hour if you would rather drive it in than post it. Either way you are told the moment it is logged, and the free diagnostic finishes two working days later.
Unsure what ought to go in the box? Ring 0800 689 0668 before you seal it, or work through the free online diagnostic and let it do the asking.