HP XP7 Owner Guide (H6F56-96006)

Array groups and RAID levels
The array group (also called parity group) is the basic unit of storage capacity for XP7 Storage.
Each array group is attached to both boards of a DKA pair over 2 SAS paths, which enables all
data drives in the array group to be accessed simultaneously by a DKA pair. Each controller rack
has two drive chassis (factory designation DKU), and each drive chassis can have up to 128
physical data drives.
The XP7 supports the following RAID levels: RAID1, RAID5, RAID6. RAID0 is not supported on the
XP7. When configured in four drive RAID5 parity groups (3D+1P), ¾ of the raw capacity is
available to store user data, and ¼ of the raw capacity is used for parity data.
RAID1. Figure 13 (page 24) illustrates a sample RAID1 (2D+2D) layout. A RAID1 (2D+2D) array
group consists of two pairs of data drives in a mirrored configuration, regardless of data drive
capacity. A RAID1 (4D+4D) group combines two RAID1 (2D+2D) groups. Data is striped to two
drives and mirrored to the other two drives. The stripe consists of two data chunks. The primary
and secondary stripes are toggled back and forth across the physical data drives for high
performance. Each data chunk consists of either eight logical tracks (mainframe) or 768 logical
blocks (open systems). A failure in a drive causes the corresponding mirrored drive to take over
for the failed drive. Although the RAID5 implementation is appropriate for many applications, the
RAID1 option can be ideal for workloads with low cache hit ratios.
NOTE: When configuring RAID1 (4D+4D), HP recommends that both RAID1 (2D+2D) groups
within a RAID1 (4D+4D) group be configured under the same DKA pair.
Figure 13 Sample RAID1 2D + 2D layout
RAID5. A RAID5 array group consists of four or eight data drives, (3D+1P) or (7D+1P). The data
is written across the four (or eight) drives in a stripe that has three (or seven) data chunks and one
parity chunk. Each chunk contains either eight logical tracks (mainframe) or 768 logical blocks
(open). The enhanced RAID5+ implementation in the XP7 minimizes the write penalty incurred by
standard RAID5 implementations by keeping write data in cache until an entire stripe can be built
and then writing the entire data stripe to the drives. The 7D+1P RAID5 increases usable capacity
and improves performance.
Figure 14 (page 25) illustrates RAID5 data stripes mapped over four physical drives. Data and
parity are striped across each of the data drives in the array group (hence the term “parity group”).
The logical devices (LDEVs) are evenly dispersed in the array group, so that the performance of
each LDEV within the array group is the same. This figure also shows the parity chunks that are
the Exclusive OR (EOR) of the data chunks. The parity and data chunks rotate after each stripe.
The total data in each stripe is either 24 logical tracks (eight tracks per chunk) for mainframe data,
or 2304 blocks (768 blocks per chunk) for open systems data. Each of these array groups can be
24 Functional and operational characteristics