A first array of disk drives overlaps with a second array of disk drives in a Redundant Array of Inexpensive Drives (RAID) system, in which the first and second arrays share at least one disk drive. A first stripe of data from a first client is stored in the first array, and a second stripe of data from a second client is stored in the second array. The shared disk drives are less than the number of drives needed to reconstruct a full stripe. Thus, in the event of a drive failure in the first array, the first client can reconstruct the first data stripe, but is never able to reconstruct the second stripe. Likewise, in the event of a drive failure in the second array, the second client can reconstruct the second data stripe, but is never able to reconstruct the first stripe.