Here's what the different RAID levels are, and what that means for redundancy, storage capacity and performance.
In "Introduction to RAID Concepts" we looked at striping, mirroring and parity. Here's how different RAID levels use them. RAID 0 uses striping only and requires at least two disks. Performance is improved compared to a single disk. The more disks in the array the greater the improvement.
Striping does not provide redundancy, if one disk fails, all data is lost. As no space is given over to redundancy the full capacity of the disks is usable. RAID 0 favors performance and capacity over redundancy. RAID 1 uses mirroring only and requires two disks. Read speed may be improved by reading in the same way as in a striped array, but not all controllers support this. Write speed is not improved, and in some cases may be slower than a single disk. Data remains available as long as one drive from a mirrored set is online. When a failed drive is replaced the array is rebuilt by copying the surviving disk.
The capacity of the array is the same as that of a single drive. RAID 1 favors redundancy over performance and capacity. RAID 5 combines striping with parity and requires at least three disks. Read speed is improved compared to a single disk, and scales with the number of disks. Write speed is also improved, but by less due to the overhead of parity calculation and storage.
The usable capacity is (N 1)*S where N is the number of drives and S is the capacity of the smallest drive. A 4 disk RAID 5 array of 1TB drives has a usable capacity of 3TB. If a single drive fails no data is lost, but performance may be reduced until the disk is replaced, and the array rebuilt. If two drives fail simultaneously all data is lost.
The more drives you have in an array, the more likely this becomes. RAID 5 sacrifices some performance and capacity to provide redundancy. RAID 6 uses striping with double parity and requires at least 4 disks. Performance is improved from a single disk, but by less than using the same number of disks in RAID 5 due to the increased parity overhead. The additional parity information allows the array to survive the simultaneous loss of any two drives but reduces available capacity.
The usable capacity is (N 2)*S. A 4 disk RAID 6 array of 1TB drives has a usable capacity of 2TB. RAID 6 makes the same compromise as RAID 5, but favors redundancy more. RAID 10 uses both striping and mirroring and requires at least 4 disks. Drives are first mirrored in RAID 0, data is then striped across the mirrors in RAID 1.
This can improve performance compared to RAID 5 or 6 as no parity calculation or storage is required. RAID 10 arrays can survive the loss of a single disk from each mirror set, but if a mirror set is lost entirely all data is lost. Usable capacity is (n/2)*s. A 4 disk RAID 10 array of 1 TB drives has a usable capacity of 2TB RAID 10 favors performance and redundancy over capacity.
Thanks for watching.
