HP Surestore NAS - Manual (Page 76)

76
RAID 1+0 provides data redundancy and good performance. However, the performance is
achieved by using a less efficient technique of storing redundant data called “mirroring.” Mirroring
maintains two sets of the data: a primary set and a backup set, or “mirror”, of the primary set.
Therefore, half of the disk space is consumed by redundant data. The data is also striped across
the disks in the array.
Figure 2 is an example showing the distribution of primary data and mirror data in a RAID 1+0
configuration. The example shows one LUN with 10 stripes, each stripe with two data segments:
n (primary segment) and n’ (“n prime” or “mirror” segment). The following data protection scheme
is demonstrated for RAID 1+0:
o If two non-adjacent disks fail, there is no data loss. Either a primary segment or
its mirror segment is still available.
o If two adjacent disks fail, the data in the entire LUN is lost. In this example, two
segments and their mirrors are lost and cannot be recovered.
Note: Disks A and E are considered as adjacent disks. The segment size is some number (x)
times 520-byte blocks.
Figure 2: RAID 1+0
RAID 5DP
RAID 5DP provides data redundancy and improves cost-efficiency by using a more efficient
method of storing redundant data. However, there is a performance penalty for each write
operation. This can impact system performance if your applications frequently update large
amounts of data. RAID 5DP provides high read throughput for small block-size requests (2 KB to
8 KB) by allowing simultaneous read operations from all of the disks in the LUN.
RAID 5DP uses two independent parity algorithms to create two parity values. The second
algorithm provides an additional means to reconstruct data, even if a second disk fails while
another disk is being rebuilt. This “double parity” scheme is referred to as “P + Q Parity.”
During a write operation, the array controller calculates “P” and “Q” parity, then distributes
(stripes) the data and parity information across the disks in the LUN. The process of calculating
and writing parity information affects the write performance of the array.
If a disk fails or becomes inaccessible, the array dynamically reconstructs all user data from the
data and parity information on the remaining disks. If another disk fails, with double parity, it can
also be recovered even while the first disk is being recovered. When a failed disk is replaced, the
array automatically rebuilds the new disk with the data that once resided on the failed disk.
Note: Until a failed disk is replaced (or a rebuild on an active spare is completed), the LUN
operates in a degraded state. The LUN must use data and parity information on the
remaining disks to recreate the contents of the failed disk, thus reducing performance.