Windows 2000 Performance Guide

RAID Disk Organizations

Going back to the original Berkeley paper published in 1988 that first enumerated RAID Levels 1 through 5, the disk industry has struggled to come up with terminology to describe different disk array architectures. A broad-based effort by the RAID Advisory Board (RAB) resulted in some much-needed standardization of the terms involved. Disk vendors seeking to establish a competitive advantage for their products often deviate from standard usage. We try to relate proprietary architectures to more generally accepted terminology, but that is not always possible.

RAID Level 0

RAID Level 0 describes simple disk striping where no additional reliability data is added to the array. The RAID 0 designation became accepted after the Berkeley RAID group defined RAID Levels 1 through 5. RAID 0 is not really RAID because there is neither redundant data or fault tolerance. The Berkeley authors accepted this classification of striping as RAID 0 because of the primacy of striping in their original formulation.

Figure 10-1 shows the organization of data in a RAID array. A single request is spread across multiple disks, allowing for parallel processing of each disk request. Because RAID 0 striping does not store any redundant error correction code data, it is the cheapest of the various disk array architectures. Striping data over multiple disks costs no more than writing that same data to individual disks. In contrast, each of the redundant disk array options described next ...